Flavonoid Productivity Optimized for Green and Red Forms of Perilla frutescens via Environmental Control Technologies in Plant Factory

Perilla frutescens (Lamiaceae) is a dietary staple in Asia. It is an abundant source of avonoids that are bioactively benecial to human health and tness.­e current popularity of plant-based consumption is being driven by the healthful benets of bioactive nutrition, and the concentration of bioactive agents found in raw plant materials is an important factor in the assessment of food quality. To test the feasibility of promoting avonoid productivity in perilla plants via environmental treatment, plant factory technology was applied to perilla plant cultivation. Apigenin (AG) and luteolin (LT) are two of the most potent anticarcinogenic avonoids in perilla, and these are also found in many vegetables and fruits. Quantitative analysis of AG and LTwas conducted on plants cultivated under nine environmental forms of treatment imposed by three levels of light intensity (100, 200, and 300 μmol·m·s) combined with three levels of nutrient-solution concentration (1.0, 2.0, and 3.0 dS·m) for hydroculture. ­e contents of AG in green and red perilla plant were increased by high nutrient-solution levels under the same light intensity. In green perilla, the highest concentration of AG (8.50 μg·g) was obtained under treatment of the highest level of nutrient-solution (3.0 dS·m) and 200 μmol·m·s of light intensity, whereas in red perilla, the highest concentration of AG (6.38 μg·g) was achieved from the highest levels of both of these forms of treatment (300 μmol·m·s and 3.0 dS·m). ­e increase in AG content per plant between the lowest and the highest levels was recorded by 6.4-fold and 8.6-fold in green and red perilla, respectively.­e behavior of LTconcentration diered between green and red forms of perilla. LTconcentration in red perilla was enhanced under nutrient deciency (1.0 dS·m) and aected by light intensity. Dierent responses were observed in the accumulations of AG and LT in red and green perilla during treatments, and this phenomenon was discussed in terms of biosynthetic pathways that involve the expressions of phenylpropanoids and anthocyanins. ­e total yield of avonoids (AG and LT) was improved with the optimization of those forms of treatment, with the best total yields: 33.9mg·plant in green Perilla; 10.0mg·plant in red perilla, and a 4.9-fold and a 5.4-fold increase was recorded in green and red perilla, respectively. ­is study revealed that avone biosynthesis and accumulation in perilla plants could be optimized via environmental control technologies, and this approach could be applicable to leafy vegetables with bioactive nutrition to produce a stable industrial supply of high avonoid content.


Introduction
Secondary metabolites (SMs) in plants are an important part of food nutrition, and their content could be used to determine food quality. SMs include terpenoids, phenylpropanoids, avonoids, and alkaloids that are essential food ingredients that determine avor, color, and taste in grains, vegetables, and fruits [1]. In addition, since the bene cial bioactivities to human health from compounds such as antioxidants have become well known, many SMs are being recognized as having medicinal e ects on both the human body and mind [2,3].
Recently, the intake of foods rich in bioactive SMs has gained in popularity worldwide. However, maintaining stable supplies of raw plant materials that can be used to produce foods high in nutritional value remains problematic due to the lack of an effective method to control the SM content in plant cultivation. Both the production and accumulation of SMs are sensitive to environmental conditions [4,5]. Generally, plants grow rapidly in less stressful environments and produce better yields in biomass. On the other hand, both biotic and abiotic factors of stress induce the formation of SMs, such as phenolic compounds [6]. Meanwhile, a few attempts have been made to evaluate the quality of plants grown under different environments by detailing the stable expressions of the secondary metabolisms and concentrations of bioactive SMs that are expected to be included in plant-based foods as an additional health promoting component [7].
A plant factory is a food production system that can supply stable plant materials year-round with no influence from climate change [8], which has proven to be effective for medicinal plant production since the environmental conditions can be fully controlled. Owing to high levels of bacterial and insect control, no pesticides are needed for the plants produced in these systems. erefore, plant factories can produce safe food with a desired level of quality when the optimal environmental condition can be determined for a particular crop. Regulating environmental factors such as light, temperature, and water has a significant impact on plant metabolic and synthetic pathways and enhances the SM content. It is well known that concentrations of SMs in plant are varying with its growth conditions. Providing high stressful environment to plants may promote accumulation of SMs; however, higher stress levels usually meanwhile result in lower yields. In the practical production in plant factories, it is greatly important to keep a balance between biomass yield and SMs concentration to maximize economic benefits. erefore, quantifying the optimal stress level from a single environmental factor or from an interaction of multiple environmental factors becomes crucial for actual production of medicinal plants in a commercial plant factory. Unlike plant production in outdoor condition, a plant factory can precisely tune its environmental factors such as light intensity, light spectrum, nutrient solution level, temperature, air flow rate, etc. without limitation of locations. If the treatments and their effects on SMs synthesis are quantified, this technique can be directly applied into commercial plant factories all over the world not only theoretically but also practically.
Perilla frutescens (Lamiaceae) is widely used as a culinary herb in Asia, particularly in Japan. e color and flavor of perilla are familiar parts of traditional Japanese cuisine. In Japan, the extract of perilla is used to add color and flavor to pickled vegetables and Japanese plums. In Japan, Korea, and India, people commonly use perilla leaves in the preparation of raw fish and shell-fish to reduce the odor, and it is added to grilled red meat to add flavor [9]. As a medicinal herb for crude drugs, the extract of perilla is used for clinical applications, being listed in the pharmacopoeia in Japan [10], Republic of Korea [11], and in People's Republic of China [12].
Apigenin (AG) and luteolin (LT) are the main flavonoids present in perilla and are also contained in many vegetables and fruits: celery, parsley, and onions for AG; red peppers, lettuce, berries, and onions for LT. AG, and LT are known as antioxidant, anti-inflammatory, and anticarcinogenic agents and, as such, they have attracted researchers' attention due to their potential as chemopreventive agents [1,[13][14][15]. Because of that, AG and LT have been candidate compounds for many pharmaceuticals or nutraceuticals from plantderived dietary agents for the development of new therapies [1,[13][14][15]. For medical use, however, it must be stated that many flavonoids including AG and LT have hermetic effects that have the potential to make them toxic, so that dose adjustment is necessary to achieve safety and efficacy [15]. us, the concentrations of AG and LT in plant-based foods are important and must be controlled by meeting the challenges of technology that can supply these raw materials in good quality.
In a previous study, we found that rosmarinic acid (RA), a major phenylpropanoid compound in herbs of Lamiaceae family, is highly increased in P. frutescens grown under uptake stress created by a nutrient-limited condition combined with high light intensity (LI) in plant factories, while maintaining a constant concentration of perillaldehyde (PA), a main terpenoid found in perilla essential oils [16]. Given the differences that exist between terpenoids and phenylpropanoids with respect to biosynthetic pathways and accumulation, there likely are no synchronous increases in PA and RA. However, since the biosyntheses of flavonoids share the biosynthetic pathway with phenylpropanoids, it remains unclear how flavonoids might react to an environment that affects the production of phenylpropanoids.
is should be clarified in order to establish quality control for raw plant materials according to the concentrations of compounds that are a rich source of food nutrition. erefore, in the present study, we investigated the quantitative production of flavonoids in a perilla plant via the gradual introduction of stresses commonly encountered in cultivation and compared the quantitative expressions of AG and LT. In a plant factory with artificial lighting, perilla plants were cultivated under nine different levels of environmental treatment created by a combination of light intensity and nutrient concentration in hydroponics. Quantitative chemical analysis via liquid chromatographymass spectrometry (LC-MS) was conducted to measure the concentrations of AG and LT as two major flavonoids in perilla plants grown under different growth conditions.

Treatments.
ree weeks after sowing, red perilla and green perilla seedlings were transplanted into a walk-in type plant factory (2.9 m × 2.0 m × 2.3 m in LWH) and subjected to three LI levels (100, 200, and 300 µmol·m −2 ·s −1 ) with a photoperiod of 16 h per day supplied by cool white fluorescent lamps and three EC levels (1.0, 2.0, and 3.0 dS·m −1 ) for five weeks. e LI was measured at the surface of the rockwool cubes using a light meter (LI-250A; Li-Cor Inc., Lincoln,NE, USA) before placing the plants. e experiment was set up in a 3 × 3 full factorial in split plot design with LI as the main plot and EC levels as subplot, and each treatment contained 18 plants. ree nutrient solution tanks (EC of 1.0, 2.0, and 3.0 dS·m −1 for each) were prepared beside the cultivation system. e plants were irrigated every 2 days from the bottom using fresh nutrient solution from each tank. e overflowed nutrient solution was discarded after each rockwool cube was saturated. Air temperature, relative humidity, and CO 2 concentration were set at 23/20°C (light/ dark periods), 60-80%, and 1000 µmol·mol −1 , respectively.

Extraction.
Extraction from perilla leaves was conducted according to Japanese Pharmacopoeia [10] and methods described in a previous article [16]. e leaves were sampled and dried at 30°C for 2 weeks. e dried perilla leaves were ground to powder and filtered through a sieve. A 10.00-10.50 mg sample was weighed accurately and transferred to a 1.5 mL tube. Methanol (1 mL) was added, mixed for 10 min at 2,000 rpm and 15°C using an Eppendorf ermoMixer (Hamburg, Germany), and centrifuged for 5 min. To the residue, methanol (1 mL × 2) was added, and the same extract manner was performed twice. e extracts (about 3 mL) were combined and transferred to a 5 mL volumetric flask and diluted with methanol to a 5 mL total volume. e solution was filtered through Agilent 0.2 µm nylon syringe filters (Agilent Technologies Inc., Palo Alto, CA, USA) to prepare the samples for LC-MS.

Contents of AG and LT. LC-MS analysis of AG and LT
was conducted according to a method described by Nishimura et al. [17] with some modifications. A Shimadzu LC-20A prominence system (HPLC) with a SIL-20AC autosampler and a LCMS-2020 mass spectrometer (MS) equipped with an electrospray ionization (ESI) source operating in negative mode were used for identification and quantification of AG and LT by chromatographic data processed using LabSolutions software (Shimadzu, Kyoto, Japan). e HPLC conditions for AG and LT were XBridge BEH C18 column (3.5 µm, 2.1 × 150 mm, Waters, MA, USA); temperature, 40°C; flow rate, 0.2 mL/min; run time, 15 min; mobile phase, 30% acetonitrile/0.1% formic acid; and injection volume, 1 µL. e eluent was passed to the electrospray source. A capillary voltage of 3.5 kV was used in the negative ion mode. Nitrogen was used as drying gas with a flow rate of 15 L·min −1 and as nebulizing gas with a flow rate of 1.5 L·min −1 . e desolvation line temperature was set at 250°C. e ion trap was operated in full-scan mode from m/z 50 to 1000 and selected-ion monitoring (SIM) mode with m/z 269 for a molecular ion [M-H] − of AG and m/z 285 for a molecular ion [M-H] − of LT. e standards of AG and LT were dissolved in a little of THF and then diluted with methanol to prepare the standard solutions before use for identification and quantification. AG and LT contents per leaf dry weight (hereafter, AG and LT concentration) were estimated by dividing the AG and LT contents in samples by sample weight. e AG and LT standards were obtained from Tokyo Chemical Industry Co., Ltd. (Tokyo, Japan). Acetonitrile (HPLC grade) was obtained from Sigma-Aldrich Japan (Tokyo, Japan). Formic acid was purchased from Kanto Chemical Co., Inc. (Tokyo, Japan).

Statistical Analysis.
All measurements were repeated three times for each sample. Five to six plants were sampled from each treatment to determine concentrations. e means of the treatment data were subjected to analysis of variance for comparison via Tukey's test using SPSS statistical software (IBM SPSS Statistics, Version 25.0. Armonk, NY: IBM Corp.). A P value <0.05 was considered significant.

Concentrations of AG and LT in Green and Red Perilla.
Concentrations of AG and LT in green perilla are shown in Figures 1(a) and 1(b), respectively, as the content per unit of leaf dry weight. e concentration of AG was decreased under an EC of 1.0 dS·m −1 using the same LI, and it tended to increase with increases in the EC under LIs of 100 and 200 µmol·m −2 ·s −1 (Figure 1(a)). e concentration of AG was the lowest (2.95 µg·g −1 ) under an EC of 1.0 dS·m −1 and a LI of 200 µmol·m −2 ·s −1 , but it was increased to the highest level (8.50 µg·g −1 ) of concentration (2.9-fold increase) under an EC of 3.0 dS·m −1 and LI of 200 µmol·m −2 ·s −1 . LI had little effect on the AG concentration under the same EC, with the exception of a comparison between LIs of 200 and 300 µmol·m −2 ·s −1 under an EC of 3.0 dS·m −1 . e resultant LT concentration showed no significant differences as a result of treatment and was affected neither by LI nor by EC (Figure 1(b)).
Concentrations of AG and LT in red perilla are shown in Figures 1(c) and 1(d), respectively, as the content per unit of leaf dry weight. e concentrations of AG tended to decrease with decreases in EC under the same LI, but LI showed little effect when EC treatment was unchanged. e lowest values

AG and LT Content per Plant in Green and Red Perilla.
e contents of AG and LT per plant in green perilla are shown in Figures 2(a) and 2(b), respectively. e highest EC

Flavonoid Production in Green and Red Perilla.
Flavonoid biosynthesis has been studied extensively in different plant species [18]. Both AG and LT have a flavone in their chemical structure, and enzymes that work as key biocatalysts to synthesize the core structure of flavonoids such as flavone, flavonol, flavanone, and anthocyanidin. RA is an ester of two C6-C3 molecules called phenylpropanoid, Values are mean ± standard error (n � 5 − 6). Different letters indicate significant differences between at P < 0.05, as determined by Tukey's test. and enzymes involved in the biosynthesis of RA have also been identified in Coleus blumei L. [19]. Recently, gene expression in P. frutescens associated with the biosynthetic pathways of flavonoids and phenylpropanoids has been revealed in studies using transcriptome analysis [20]. us, based on the knowledge above, we have proposed the biosynthetic pathways relevant to RA, AG, and LT, which are present in perilla, as shown in Figure 3. e pathway to flavones, AG and LT, includes the general biosynthesis of phenylpropanoids derived from L-phenylalanine to produce 4-coumaroyl-CoA, which is a crucial precursor for the biosynthesis of flavones and RA [19]. e substrate 4coumaroyl-CoA is one of two used for esterification, during which it reacts with 4-hydroxyphenyllactic acid in a step that is important in the biosynthesis of RA.
at fact suggests that producing molecules of AG, LT, and RA requires the use of an equivalent of 4-coumaroyl-CoA molecules as a biosynthetic precursor and also suggests the need to activate the enzymes involved in the phenylpropanoid pathway.
Our results showed that the effect of EC levels on the accumulation of AG and LT was quite different. e concentration of AG was highest for an EC of 3.0 dS·m −1 , whereas that of the LT was highest for an EC of 1.0 dS·m −1 . In addition, the AG concentration was lowest under an EC of 1.0 dS·m −1 , whereas that of the LT was lowest under an EC of 3.0 dS·m −1 . ese opposition responses between AG and LT could reflect their competitive relationship in the sharing of naringenin, which resides in the connection between the pathways to AG and anthocyanins (Figure 3). In our previous report, the concentration of RA was the highest under nutrient-limited conditions (EC of 1.0 dS·m −1 ) and decreased with the highest EC level (EC of 3.0 dS·m −1 ) in green and red perilla [16]. Interestingly, the similar responses between LT and RA were observed in red perilla, so that the LT concentration was enhanced in red perilla grown under nutrient-limited stress. e effect of EC on LT accumulation differed between green and red perilla. In green perilla, the results of LT concentration showed neither a change nor a trend, but the LT content per plant had a wide range of figures depending on the environment. e yield of LT was promoted under high levels of both EC and LI in accordance with the biomass yield (Figures 1(b) and 2(b)). In red perilla, however, the concentration of LT was enhanced under an EC of 1.0 dS·m −1 , which stood in opposition to the biomass trend (Figures 1(d) and 2(d)). is suggests that the LTmetabolism in red perilla overcame the negative effect of EC during growth when LI reached at least 200 µmol·m −2 ·s −1 , which caused an increase in the LT yield against a decrease in the biomass yield but only at this level of LI. With LI values of 100 and 300 µmol·m −2 ·s −1 , the yield of LT showed a clear dependence on the biomass yield.
We compared AG and LT accumulation in green and red perilla (Figures 4 and 5). In red perilla, LT accumulation was always higher than that of AG, whereas in green perilla Values are mean ± standard error (n � 5 − 6). Different letters indicate significant differences between at P < 0.05, as determined by Tukey's test.
AG was higher than LT except for three treatments (Figure 4). LT molecules have a hydroxyl group at the 3′ position on the avone skeleton, and a crucial enzyme that introduces the 3′-hydroxyl group on the corresponding precursor has been identi ed as avone 3′-hydroxylase (F3′H) (Figure 3) one of the most important catalysts in anthocyanin synthesis in Arabidopsis [20], the accumulation of LT that was higher than that of AG in red perilla, presumably caused by the higher transformation activity of F3′H in anthocyanin biosynthesis in red plants compared with that in green plants. To clarify this interaction, we tried to trace biosynthetic precursors: naringenin chalcone, naringenin, kaempferol, quercetin, and eriodictyol by quantitative analysis using LC-MS, but none of them detected from extract of perilla dry leaves. It means that those molecules were not accumulated enough in a detective level on their relative pathway. en, we though that those precursors are transformed quickly and used for biosynthesis of AG, LT, and anthocyanins. e total concentrations of AG and LT in green perilla were the lowest under a LI of 200 µmol·m −2 ·s −1 and an EC of 1.0 dS·m −1 , but these were increased to the highest levels (2.2-fold increase) of concentration under LI of 200 µmol·m −2 ·s −1 and an EC of 3.0 dS·m −1 (Figure 5(a)). In red perilla, however, concentrations were the lowest under LI of 200 µmol·m −2 ·s −1 and an EC of 3.0 dS·m −1 and increased to the highest levels (1.4-fold increase) under LI of 300 µmol·m −2 ·s −1 and an EC of 1.0 dS·m −1 (Figure 5(c)). In other words, the range of total concentration was narrower in red plants compared with green plants, which indicated that red plants could withstand a greater environmental impact on flavone accumulation, which could result in greater uniformity of food quality. e total yield of AG and LT in green perilla was the lowest under LI of 200 µmol·m −2 ·s −1 and an EC of 1.0 dS·m −1 and increased to the highest yield (a 4.9-fold increase) under LI of 200 µmol·m −2 ·s −1 and an EC of 3.0 dS·m −1 (Figure 5(b)). For red perilla, the total yield of AG and LT was the lowest under LI of 100 µmol·m −2 ·s −1 and an EC of 1.0 dS·m −1 and increased to the highest yield (a 5.4-fold increase) under LI of 300 µmol·m −2 ·s −1 and an EC of 3.0 dS·m −1 (Figure 5(d)). Because green plants obtained larger biomass yields than red plants, they had the best total yield of AG and LT (33.9 mg·plant −1 ; LI: 200 µmol·m −2 ·s −1 ; EC: 3.0·dS·m −1 ) that were triple more than that of red plants (10.0 mg·plant −1 ; LI: 300 µmol·m −2 ·s −1 ; EC: 3.0 dS·m −1 ) under application of the best treatments for both (Figures 5(b) and 5(d)), in which the total concentrations of AG and LT in green and red plants were almost the same, 14.6 µg·g −1 and 13.3 µg·g −1 , respectively ( Figures 5(a) and 5(c)).

Conclusions
We examined the bioactive flavonoids that appear in green and red forms of perilla plants. e results indicated that LI and EC were the factors that most affected flavonoid productivity in both concentration and yield of apigenin and luteolin.
erefore, environmental treatments that were regulated by those factors had a remarkable impact on  flavone biosynthesis and accumulation, and their use is proposed as an effective method for the cultivation of green and red perilla plants. ese findings could be applicable to an optimization of cultivation control in the production of foods composed of leafy vegetables containing bioactive compounds.

Data Availability
e data used to support the findings of this study are included within the article.

Conflicts of Interest
e authors declare that there are no conflicts of interest regarding the publication of this paper.