Establishment and optimization of a hemp (Cannabis sativa L.) agroinfiltration system for gene expression and silencing studies

Industrial hemp (Cannabis sativa L.) is a high-yielding annual crop primarily grown for fiber, seeds, and oil. Due to the phytochemical composition of hemp, there has been an increased interest in the market for nutraceuticals and dietary supplements for human health. Recent omics analysis has led to the elucidation of hemp candidate genes involved in the syntheses of specialized metabolites. However, a detailed study of these genes has not been undertaken due to the lack of a stable transformation system. We report for the first time an agroinfiltration system in hemp utilizing vacuum infiltration, which is an alternative method to stable transformation. A combination of 0.015% Silwett L-77, 5 mM ascorbic acid, and thirty second sonication followed by a 10-minute vacuum treatment resulted in the highest β-glucuronidase expression in the leaf, male and female flowers, stem, and root tissues. The phytoene desaturase gene was silenced with a transient hairpin RNA expression, resulting in an albino phenotype in the leaves and the male and female flowers. This agroinfiltration system would be useful for overexpression and silencing studies of target genes to regulate the yield of specialized metabolites in hemp.

Industrial hemp (Cannabis. sativa L.) is a diploid (2n = 20) and typically dioecious plant primarily grown for fiber, seeds, and oil. Recently, this crop has been exploited for medicinal compounds as it produces 565 secondary metabolites, such as cannabinoids, terpenes and other phenolic compounds 1 . Unlike drug-type Cannabis (marijuana) that produces high levels of the psychoactive component Δ9-tetrahydrocannabinol (THC), hemp preferentially produces cannabidiol (CBD) and low amounts of THC (0.3% compared with marijuana). In addition, a total of 25,000 hemp products are currently produced and sold on the market, for both scientific and industrial purposes 2 .
Cannabinoids are terpenophenolic compounds that include more than 100 chemical compounds, such as THC and CBD 2 . These molecules have pharmacological properties in the treatment of pain, mood disorders, diabetes, cancer, and neurodegenerative and inflammatory diseases 3,4 . The biosynthesis of these molecules occurs in the glandular trichomes located predominantly on the flowers and leaves of C. sativa 1 . The cannabinoid pathway is initiated by the synthesis of olivetolic acid (OA) originating from the primary metabolite precursor via hexanoyl-CoA by tetraketide synthase (type III polyketide synthase) and olivetolic acid cyclase 5 . Together with geranyldiphosphate (GPP) from the plastidal 2-C-methyl-D-erythritol-4-phosphate pathway, OA is converted into cannabigerolic acid (CBGA) by CBGA synthase. This cannabinoid pathway concludes with the syntheses of tetrahydrocannabinolic acid (THCA) and cannabidiolic acid (CBDA) by THCA synthase (THCAS) and CBDA synthase (CBDAS), respectively 6 . THCAS and CBDAS are converted to THC and CBD, respectively, by nonenzymatic reactions.
Terpenes have anxiolytic, antibacterial, anti-inflammatory, and sedative effects on human diseases 7 . For example, the sesquiterpene β-caryophyllene interacts with mammalian cannabinoid receptors to reduce inflammation 8 . The terpene biosynthesis pathway in C. sativa includes two pathways: the plastidial methylerythritol phosphate (MEP) pathways, which leads to the production of monoterpenes, and the cytosolic mevalonate (MEV) pathway, which leads to the production of sesquiterpenes 1 . Seven enzymes have been shown to produce two terpene precursors: GPP and farnesyl diphosphate (FPP). Furthermore, the nine terpene synthase genes that catalyze terpene production from GPP and FPP have been identified 9 .
Knowledge of these secondary metabolite synthesis genes will help facilitate genetic improvements to obtain desirable metabolite profiles. However, these candidate genes were proposed based on genomic, transcriptomic and proteomic analyses and remain to be studied. Towards this goal, stable transformation is ideal; however, a hemp transformation protocol has not yet been developed due to the low shoot regeneration rate 10,11 .
Agroinfiltration is a prominent methodology for transiently expressing or silencing genes of interest in an efficient manner. This method can be used to produce vaccines and enzymes for industrial use as an alternative to stable transformation 12 . Agroinfiltration studies have also provided insight into various biological processes, such as promoter function, gene expression, subcellular protein localization, protein-protein interactions, and metabolism 13 . To date, this agroinfiltration system has been developed and utilized in both 1) model plant species such as tobacco, Nicotiana benthamiana, Arabidopsis, tomato and 2) agronomic crops such as rice, soybean, onion, flax, cowpea, grapevine, rose, and cacao 14 . Depending on the plant species, the agroinfiltration efficiency varies, resulting in high expression in plants such as N. benthamiana and tobacco or low expression in a wide range of plant species, including hemp. For infection of the plant leaf cells, Agrobacterium suspension is injected into the leaf parenchyma, but in many cases, agroinfiltration is not successful because the morphology of the leaf and particularly the structure of the thick epidermis do not allow the bacterial culture to enter the parenchyma, consequently resulting in a low gene expression efficiency 15 .
In hemp, a hairy root culture system for transient gene expression assays was developed and resulted in slight β-glucuronidase (GUS) positive staining using Agrobacterium tumefaciens and A. rhizogenes 16 . However, pharmaceutical components are not abundant in the C. sativa root 17 , and therefore, this system is not appropriate for analyses of the genes encoding enzymes that synthesize specialized metabolites.
To our knowledge, this article constitutes the first report of C. sativa agroinfiltration with a high efficacy in the aerial parts of the plant where pharmaceutical components are highly synthesized, transported and accumulated. The agroinfiltration method was optimized by a combination of physical and chemical treatments with different Agrobacterium strains and hemp cultivars. This optimized system had a high efficiency for exogeneous gene expression and silencing of endogenous genes in different hemp tissue/organs.

Results
Optimization of agroinfiltration. The GUS gene (uidA), which encodes the enzyme GUS, is one of the most widely used reporter genes in plant molecular biology 18 . For analysis of the efficiency of agroinfiltration, the uidA gene was cloned into the binary vector pEarleyGate 101 and expressed under the control of the cauliflower mosaic virus (CaMV) 35 S promoter (Fig. 1a). This cassette was transformed into Agrobacterium GV3101. Further agroinfiltration was carried out, and the tissues underwent histochemical GUS staining.
Agroinfiltration was carried out via vacuum infiltration to test the different tissues of the plant. The mature leaf, male flower, female flower, stem and root tissues were excised from the hemp plants two months after seed germination. In a histochemical GUS assay, only the male flower in all the tissues tested showed weak GUS staining. Then, various parameters such as chemical additives, sonication, different vacuum time, hemp cultivars and Agrobacterium strains were optimized using the male flower to create an optimal agroinfiltration protocol.  Silwett L-77 and Pluronic F-68 are surfactants that can reduce the surface tension and enhance the entry of bacteria into plant tissues 19 . The addition of Silwett L-77 at 0.015% drastically increased the GUS expression to approximately three times that of the negative control plant (0% Silwett L-77) (Fig. 2a). The percentage of the GUS-stained male flowers was higher in the 0.015% Silwett L-77 group (11.1%) than the negative control group (2.2%). This effect was significant for both relative GUS expression and the percentage of GUS-stained female flowers with 0.015% Silwett L-77. Therefore, Silwett L-77 at a final concentration of 0.015% was added to the agroinfiltration media for all subsequent agroinfiltration experiments. Addition of Pluronic F-68 led to a 25% increase in the relative GUS expression and a 2.5-fold increase in the percentage of GUS-stained female flowers (24.4%) at a concentration of 0.05%, while 0.5% did not display any positive effect (Fig. 2b), presumably due to tissue damage by the high concentration of this surfactant (Fig. 2b).
Ascorbic acid and PVP were tested because of their roles as antioxidants that scavenge the excess reactive oxygen species (ROS) produced by Agrobacterium infection. Ascorbic acid concentrations correlated with increased relative GUS expression (1.3-fold increase compared to the negative control) and a higher percentage of GUS-stained female flowers (18.9%) compared with the negative control (10%), with a final concentration of 5 mM (Fig. 2c). No significant increases in the GUS level were observed in the plants treated with PVP (Fig. 2d).

Vacuum and sonication time.
Vacuum infiltration for 5 minutes resulted in a higher relative GUS expression (5.5-fold) and percentage of GUS-stained flower buds (17.7%) than the non-vacuum treatment (Fig. 3a). Increasing the vacuum time to 10 minutes and 15 minutes and three iterations of 5 minutes resulted in a higher relative GUS expression than that with 5 minutes of vacuum treatment, but there was no significant effect on the percentage of GUS-stained flower buds (Fig. 3a).
Exposure of the plant to sonication in the presence of Agrobacterium produces many microwounds across the tissue, which allows the Agrobacterium to penetrate deeper and more completely throughout the tissue 15 . Sonication for 30 seconds before vacuum infiltration had a significant effect on the nonsonicated male flowers for both relative GUS expression (1.2-fold) and the percentage of GUS-stained male flowers (22.2%; Fig. 3b). A combination of different vacuum and sonication times with ascorbic acid was tested. Ten minutes of vacuum infiltration after 30 seconds of sonication together with the addition of 0.05% Pluronic F-68 and 5 mM of ascorbic acid resulted in GUS staining of more than 50% of the male flowers and the highest relative GUS expression (3-fold) and GUS transcript level (Fig. 3c,d).
Agrobacterium strain and hemp cultivars. Three disarmed Agrobacterium strains, EHA105, LBA4404 and GV3101 containing the pEarlyGate 101-uidA binary vector, were used to perform the histochemical GUS assays. Among the three Agrobacterium strains, GV3101 demonstrated an approximately 1.7-fold increase in GUS www.nature.com/scientificreports www.nature.com/scientificreports/ expression compared to the other two strains (Fig. 4a); therefore, GV3101 was used to evaluate the effect of hemp cultivars on the transient GUS expression.
Since the structures of the male flowers are different between monoecious and dioecious cultivars, it is not possible to compare the transient GUS expression in male flowers. Therefore, mature leaves were used to evaluate the effect of the hemp cultivars on the transient GUS expression. The dioecious CRS-1 cultivar had the highest relative GUS expression (Fig. 4b). Santhica 27, CFX-2 and Futura 75 exhibited approximately 80% of the activity compared to CRS-1. Fedora17, USO31 and Ferimon showed approximately 60% relative GUS expression compared to CRS-1. Felina 32 showed the lowest GUS expression among all eight cultivars tested (Fig. 4b).
The pEarleyGate 101-eGFP vector (Fig. 1b) was also transiently expressed via a GV3101 strain of Agrobacterium. GFP fluorescence was detected in the hemp leaf discs (Fig. 6d), pollen sac clusters (Fig. 6f), anthers and sepals (Fig. 6h), filaments (Fig. 6j), capitate-stalked trichomes (Fig. 6l) and roots in the hemp seedlings (Fig. 6n). The hemp leaf discs filtrated by the empty pEarley vector did not show any GFP fluorescence (Fig. 6b). The bright field image of a leaf disc that was treated with the empty vector pEarleyGate 101 is shown in Fig. 6a. The bright field images of the leaf discs, pollen sacs, anthers and sepals, filaments, capitate-stalked trichomes and roots that were treated with pEarleyGate 101-eGFP are shown in Fig. 6c,e,g,i,k,  www.nature.com/scientificreports www.nature.com/scientificreports/ fragment to form the hairpin RNA. This silencing vector was transformed into the GV3101 strain and used for hemp agroinfiltration. Four days after inoculation, an albino phenotype was observed in the mature leaves, male flowers and female flowers (Fig. 7a). qPCR analysis demonstrated that the CsPDS gene expression was sharply decreased to less than 20% compared with that of the control tissues that were inoculated by GV3101 harboring the empty vector pK7GWIWG2(I) (Fig. 7b).

Discussion
Vacuum agroinfiltration was attempted in mature leaves, male flowers, and female flowers, but only the male flowers exhibited a low transient GUS expression. Then, we used male flowers to optimize the hemp agroinfiltration. A concentration of 0.015% Silwett L-77 drastically increased the transient GUS expression and resulted in three times higher relative GUS expression than that without Silwett L-77, demonstrating that Silwett L-77 is www.nature.com/scientificreports www.nature.com/scientificreports/ a critical factor for successful hemp vacuum agroinfiltration (Fig. 2a). Thus, this concentration was used for all subsequent agroinfiltration experiments. Pluronic F-68 at 0.05% also increased the GUS activity by approximately 20% (Fig. 2b). The efficiency of both surfactants on the GUS expression in the hemp male flower (Fig. 2a,b) proved that the addition of these surfactants effectively increased the agroinfiltration efficiency in hemp as well as various crops 22,23 .
In plant cells, agroinfiltration triggers an oxidative burst response and induces the accumulation of ROS that injures plant cells and leads to necrosis 24 . Excess levels of ROS were reported to reduce the ability for Agrobacterium to colonize plant cells and transfer T-DNA into the nuclei 25 . For scavenging of the excess ROS, two antioxidant compounds were tested: ascorbic acid and PVP 26,27 . A 5 mM concentration of ascorbic acid had a significant and positive effect on transient GUS activity and increased the relative GUS expression by 50% (Fig. 2c), whereas PVP did not significantly improve the efficiency of agroinfiltration (Fig. 2d). These results are consistent with a N. benthamiana infiltration study previously reported by Norkunas et al. (2018) 22 .
For vacuum agroinfiltration, plant tissue was submerged into the suspension of Agrobacterium harboring a binary vector. Application of a vacuum results in evacuation of the air in the interstitial space from the submerged www.nature.com/scientificreports www.nature.com/scientificreports/ plant tissue through the stomata. The suspension of Agrobacterium enters the plant tissue to replace the evacuated gases once the vacuum is broken and the internal pressure is increased. Long periods of vacuum time lead to damaged plant tissues and short vacuum times result in insufficient air evacuation and low agroinfiltration efficiency 14 . In this study, both 10 minutes and 15 minutes as well as 3 iterations of 5 minutes of the vacuum treatment resulted in approximately seven times higher relative GUS expression than that of the control (non-vacuum) male flower. In statistical tests, these values were significantly higher than those after shorter periods of vacuum application for both relative GUS expression and the percentage of GUS-stained flower buds (Fig. 3a).
Sonication leads to the formation of microwounds within the plant tissue, allowing Agrobacterium to effectively infect the plant cells 28 . Since this treatment produces more entry points for Agrobacterium to enter into the plant cell, sonication facilitates agroinfiltration in recalcitrant plants. We performed tests to determine the best sonication time for hemp and found that 30 seconds of sonication before vacuum infiltration led to 20% higher relative GUS expression than that without sonication (Fig. 3b).
Thus, Silwett L-77, Pluronic F-68, ascorbic acid, vacuum and sonication all proved to be effective for hemp agroinfiltration. To further optimize the agroinfiltration efficiency, we tested a combination of Silwett L-77, Pluronic F-68, ascorbic acid, vacuum and sonication. As a consequence, ten minutes of vacuum infiltration after 30 seconds of sonication together with the addition of 0.015% Silwett L-77, 0.05% Pluronic F-68 and 5 mM of ascorbic acid resulted in 2.5-fold higher uidA gene expression and staining of more than 50% of the male flowers than those of the Silwett L-77 only group (Fig. 3c). This result indicates the synergistic effect of Silwett L-77, Pluronic F-68, ascorbic acid and acetosyringone.
The genetic background of the Agrobacterium can influence the T-DNA transfer into plant cellular nuclei, and some Agrobacterium strains are more virulent than others based on the target species 29,30 . In this study, different transformation rates were observed for three avirulent strains. To the best of our knowledge, there are few reports about appropriate Agrobacterium strains for hemp due to the limited tissue culture and transformation studies. The GV3101 strain demonstrated significantly higher GUS expression (1.7-fold) than the other two strains (Fig. 4a), and it therefore should be the optimal choice for C. sativa. Ti-plasmids are classified into several types based on the opines; EHA105, LBA4404 and GV3101 are grouped into the succinamopine, octopine and nopaline types, respectively 25 . Since GV3101 showed high transient GUS expression, octopine-type strains might have better compatibility with C. sativa, and other Agrobacterium strains belonging to this type, such as C58C1, GV3100, GV3850, A136, and EHA 101, should be tested for hemp agroinfiltration.
C. sativa is a dioecious plant, but to overcome the unsynchronized maturity of male and female plants and the problem with mechanized harvesting, genetically monoecious diploid (2n = 20) hemp cultivars have been developed 31 . The dioecious cultivar CRS-1 exhibited significantly higher relative GUS expression than seven other varieties. Among the five monoecious cultivars, Santhica 27 presented the highest relative GUS expression, and therefore this cultivar will be used to optimize the agroinfiltration protocol for future studies due to the agronomical advantages of monoecious cultivars.
Interestingly, this optimized agroinfiltration protocol was applied to various hemp tissue/organs, including hemp female flowers, and resulted in high GUS and GFP expression (Figs. 5 and 6).
Recently, the PDS gene has been used to examine transient gene silencing systems in both monocots and dicots 32,33 . A mutant of this gene, which is a single copy gene in most plants, induces an albino phenotype and is easily detected through visual observations due to the reduction of chlorophyll content 34 . Transient expression of the PDS silencing vector decreased the PDS mRNA level to 20% compared with that of the mock treatment and led to an albino phenotype in the leaves, male flowers and female flowers (Fig. 7). These results revealed that the optimized hemp agroinfiltration system reported in this study was effective for gene silencing via the transient expression of hairpin RNA.
In conclusion, we report an efficient protocol for agroinfiltration of various hemp tissues, including female flowers that contain glandular trichomes where hemp phytochemicals are produced and accumulated at high concentrations. This method is therefore a promising technique for functional genomics of hemp genes, particularly for the synthesis of cannabinoids, terpenoids, flavonoids and alkaloids. Since there are several reports regarding the production of proteins related to vaccines, pigmentation, photosynthesis and pharmaceutical components via agroinfiltration 13,35,36 , this technique can also play an important role in the metabolic engineering of hemp, for example by overexpression of the CBDAS gene and/or silencing of the THCAS gene.  (Fig. 1a,b). An empty pEarleyGate 101 vector was used as a negative control (Fig. 1c).

Plant
The pEarleyGate 101-uidA was transformed into A. tumefaciens GV3101, LBA4404 and EHA105 strains 14,15 and the pEarleyGate 101-eGFP vector was transformed into a GV3101 strain. For gene silencing, the hairpin RNA-expressing RNAi construct was prepared by inserting sense and antisense partial CsPDS fragments into the pK7GWIWG2 gateway vector (Invitrogen; Fig. 1d). A 200 bp CsPDS partial sequence (PDS 601-800) was amplified in the sense orientation from a cDNA library that was prepared using total RNA from hemp leaves. For cloning of the gene fragments into the entry vector Gateway pDONR/Zeo (Thermo Fisher Scientific, Waltham, MA, USA), primers were synthesized to include the recombination sequences attB1 and attB2. Targeted gene fragments in the entry clones were then transferred to the binary silencing vector pK7GWIWG2(I) 38 to form the hairpin RNA, which is under the control of a CaMV 35 S promotor and a CaMV 35 S terminator (Fig. 1e). The construct was evaluated by sequencing and subsequently transformed into A. tumefaciens GV3101 cells. The full-length CsPDS (PK24508) sequence information was obtained from the hemp genome gateway browser (http://genome.ccbr.utoronto.ca/cgi-bin/hgGateway).
A single colony of recombinant Agrobacterium was inoculated and grown in LB buffer. The culture was then harvested and centrifuged at 4,000 g for 10 minutes and resuspended to an OD 600 of 0.5 in agroinfiltration media containing 10 mM MES, 1x MS, 2% glucose, and 200 µM acetosyringone at a pH 5.6. The chemical additives including Silwett L-77 (Lehle seeds, Round Rock, USA), Pluronic F-68 (Gibco, NY, USA), L-Ascorbic acid (Sigma-Aldrich, St Louis, MO, USA) and polyvinylpyrrolidone (PVP) (Sigma-Aldrich) were added before vacuum infiltration.
Vacuum infiltration. Hemp tissues/organs were excised from the plants two months after seed germination, placed on a 100 ×15 mm Petri dish filled with Agrobacterium solution and placed in a vacuum chamber. The vacuum pump was turned on to decrease the pressure, and the agroinfiltration time was calculated after the vacuum reached 80 mbar. Successful infiltration required bubbles to be flowing up from the hemp tissues/cultures. After 5-15 minutes at low pressure, the release valve was opened slowly to allow entrance of Agrobacterium into the interstitial spaces of the plant tissues. The plant tissues were then washed with distilled water, transferred onto a moist filter paper in a Petri dish and placed in a growth room at 21 °C. Sonication was carried out using a 22.5 kHz XL2000 Qsonica sonicator (Qsonica LLC, Newtown, CT, USA).

Analysis of transient GUS expression.
Four days after agroinfiltration, the plants were analyzed by histochemical staining for GUS activity in the hemp tissues/organs as described by Bakshi et al. (2011) with some modifications 24 . The infiltrated plant tissues/organs were washed for 30 minutes in ice-cold 100 mM phosphate buffer (pH 7.0) and then vacuum infiltrated in 5-bromo-4-chloro-3-indolyl-beta-D-glucuronide (X-Glu) substrate solution [0.0005% X-Glu (w/v), 100 mM phosphate buffer pH 7.0, 1 mM potassium ferrocyanide, 1 mM potassium ferricyanide, 0.05% Triton X-100 (w/v)]. The GUS-stained tissues/organs were then submerged in 95% ethanol overnight and washed with sterilized water for decolorization.
The GUS-stained hemp tissues/organs were placed in a Petri dish and observed using a Nikon SMZ1500 microscope (Nikon, Tokyo, Japan) at 2x magnification with Imaging Software NIS Elements F Package (Nikon) and equipped with a DS-Ril1 camera (Nikon). The relative GUS expression was calculated with 30 pollen sacs per treatment to obtain the total expression value (stained area x color intensity) with ImageJ (Rasband 1997-2011) as previously reported 15 . Photos of the strongest GUS-stained regions in the pollen sacs were taken using bright field illumination. Prior to quantification, the color mode of the light microscope images was converted from RGB to HSB using ImageJ. A rectangular region of interest (ROI) was defined by using the "Rectangular selection" drawing tool from the ImageJ toolbar. The intensity of the GUS staining was measured in the saturation channel. The mean gray values were measured and compared among different treatments. All analyzed tissues were thoroughly cleared to correlate the color information with the degree of blue staining. In addition, the number of stained pollen sacs was counted from 10 pollen sacs to calculate the percentage of GUS staining. This calculation included three repetitions. Statistical analysis was performed using a 1-way ANOVA with Tukey's multiple comparison test (a = 0.05).
Image collection and analysis of GFP fluorescence. Four days after vacuum agroinfiltration, the hemp tissues/organs were subject to GFP fluorescence observation using a Nikon SMZ1500 microscope with Imaging Software NIS Elements F Package (Nikon) equipped with a DS-Ril1 camera (Nikon), an Intensilight C-HGFI Precentered Fiber Illuminator (Nikon) and a GFP2 filter set (Ex. 480 ± 40 nm; Em. 510 nm LP).

Quantitative real-time PCR (qPCR) analysis of the agroinfiltrated hemp plants. Real-time
PCR was performed with 5 μL of SYBR Select Master Mix (Applied Biosystems, Waltham, MA, USA) in a 10 μL total reaction mixture containing 400 nM of the gene-specific primers and 1 μL of cDNA. The following thermal program was used: 95 °C for 10 m, 40 cycles at 95 °C for 10 seconds, and 60 °C for 1 minute carried out with a CFX-96 Touch Real-Time PCR Detection System (Bio-Rad, Hercules, CA, USA). The gene expression level was calculated from the cycle threshold according to the 2 −∆∆Ct method 39 . Statistical analysis was performed using a paired t-test (a = 0.05). The PDS gene primers were designed to amplify a region upstream of the siRNA sequence (PDS 466-555). The F-box gene was used as an internal reference for hemp, and three independent biological replications were conducted. All primer sequences are listed in Supplemental Table S1.