COMT, CRTZ, and F3′H regulate glycyrrhizic acid biosynthesis in Glycyrrhiza uralensis hairy roots

Glycyrrhiza uralensis Fisch. is prescribed as one of the original plants of licorice in Chinese Pharmacopoeia. This herbal medicine possesses numerous pharmacological activities and has been used in clinic in China since ancient times. Glycyrrhizic acid (GA) is a triterpenoid compound isolated from G. uralensis. It is often used as one of the marker components for the medicinal quality of the herb. Previously, we showed that the expression levels of three genes in G. uralensis were inversely correlated with the content of GA, including the caffeic acid 3-O-methyltransferase gene (COMT), the β-carotene 3-hydroxylase gene (CRTZ), and the flavonoid 3′-monooxygenase gene (F3′H). In this study, the main aim is to determine the roles of the three genes on GA production through gene knockout and overexpression in G. uralensis hairy roots. We observed that neither knockout nor overexpression of any of the genes affected the viability of the transgenic hairy roots, indicating that these genes are not essential for survival of hairy roots. Compared with the wild type and negative control hairy roots, GA content was significantly lower in hairy roots overexpressing COMT, CRTZ, or F3′H, but higher in those with anyone of the genes knocked out. Our findings demonstrate that the three genes, COMT, CRTZ, and F3′H, all negatively regulate the GA biosynthesis.


Introduction
Glycyrrhiza uralensis Fisch. is prescribed by Chinese Pharmacopoeia as one of the original plants of licorice (National Pharmacopoeia Commission 2020). It is the most commonly-used Chinese herb and has been applied in the treatment of lung diseases (Chen et al. 2018) and diabetes ) since ancient times in China. In recent Communicated by Neelam sangwan. Zhixin Zhang, Wenwen Ding, and Ziyi Chen contributed equally to this work. years, modern pharmacological studies have yielded significant insights into the bioactive properties of this ancient Chinese herb. Additional therapeutic activities of G. uralensis have been unveiled, such as anti-inflammatory Ko et al. 2021), anti-microbial Zeng et al. 2021), antiviral (Yi et al. 2022), anticancer (Jain et al. 2022;Wen et al. 2021;Xiao et al. 2011), and immunoregulatory activities (Aipire et al. 2017;Bhattacharjee et al. 2012). Numerous bioactive compounds, including triterpenoids and flavonoids, have been isolated and identified from G. uralensis. Among them, glycyrrhizic acid (GA), a pentacyclic triterpenoid, is the most studied bioactive compound of the herb due to its excellent pharmacological activities. It has been used to treat inflammation-associated diseases in clinic (Chu et al. 2020;Xie et al. 2015). Importantly, GA is stipulated by Chinese Pharmacopoeia as the quality standard compound in G. uralensis (National Pharmacopoeia Commission 2020). Thus, a better understanding of the regulatory mechanisms of GA biosynthesis will provide guidance on improving the medicinal quality of this commonly-used herb.
Multiple pathways crisscross and participate in a delicate balance that coordinates metabolism in G. uralensis. As shown in Fig. 1, the primary metabolites derived from glycolysis flow into different pathways to produce major bioactive components in G. uralensis. Among these primary metabolites, glyceraldehyde 3-phosphate, pyruvate, and acetyl-CoA are substrates for terpenoids production. They are used to synthesize dimethylallyl diphosphate (DMAPP) and isopentenyl diphosphate (IPP), the common precursors of terpenoids, through two independent and distinct pathways, the 2-C-methyl-d-erythritol 4-phosphate (MEP) and mevalonic acid (MVA) pathways (Tian et al. 2022). Subsequently, DMAPP and IPP formed via the MVA pathway are used to produce triterpenoids, such as GA, (Liao et al. 2016), Fig. 1 Primary and secondary metabolism in G. uralensis. The primary metabolism is shown in the upper gray box and the secondary metabolism in the bottom blue box. The primary metabolites from glycolysis flow into different pathways to produce various secondary metabolites. Glyceraldehyde 3-phosphate and pyruvate shunt into the MEP pathway to produce tetraterpenoids (marked in pink), acetyl-CoA flows into the MVA pathway to produce triterpenoids (marked in blue), while phosphoenolpyruvate is the substrate of the shikimate pathway and eventually forms virious lignins (marked in green) and flavonoids (marked in orange). GA is synthesized via the MVA pathway. CRTZ, COMT, and F3′H are key enzymes involved in the tetraterpenoid, lignin, and flavonol biosynthetic pathways, respectively. These pathways compete with the MVA pathway for substrates from glycolysis. MEP, 2-C-methyl-d-erythritol 4-phospate; MVA, Mevalonic acid; DMAPP, Dimethylallyl diphosphate; IPP, Isopentenyl diphosphate; IDI, Isopentenyl diphosphate isomerase; CRTZ, β-Carotene 3-hydroxylase; F3′H, Flavonoid 3′-monooxygenase; COMT, Caffeic acid 3-O-methyltransferase while the equivalent two isomeric five-carbon compounds synthesized through the MEP pathway are supplied to generate tetraterpenoids, such as β-carotene (Banerjee and Sharkey 2014). In addition, phosphoenolpyruvate shunted into the shikimate pathway forms l-phenylalanine, which eventually produces various flavonoids and lignin through the phenlpropanoid metabolic pathway (Vogt 2010). Our previous studies demonstrated that the above pathways were all tightly associated with GA accumulation Wang et al. 2021;Yin et al. 2020;Zhang et al. 2021).These pathways compete with each other for carbon sources generated from the primary metabolism. β-carotene 3-hydroxylase (CRTZ) is a key rate-limiting enzyme involved in the synthesis of tetraterpenoids . It is a multifunctional enzyme which converts β-carotene into β-cryptoxanthin and β-cryptoxanthin into zeaxanthin (Liang et al. 2020;Davison et al. 2002;Du et al. 2010). Caffeic acid 3-O-methyltransferase (COMT) mainly takes part in the biosynthesis of monolignols. It catalyzes the methylation of multiple substrates, such as caffeic acid and caffeyl alcohol (Daly et al. 2019;Liang et al. 2022;Saluja et al. 2021), and eventually affects the composition and content of lignin (Wu et al. 2019;Xie et al. 2019;Weng et al. 2011). Flavonoid 3′-monooxygenase (F3′H) belongs to cytochrome P450 monooxygenase family and plays an essential role in the hydroxylation of flavonoids (Tanaka and Brugliera 2013;Jiang et al. 2021;Liu et al. 2022). To our knowledge, none of the three genes has been cloned from G. uralensis. Hence, the potential roles of these genes were deemed worth of study.
Hairy root culturing is a technology which combines tissue culturing with genetic transformation. Hairy roots are easily manageable and fast proliferating (Wawrosch and Zotchev 2021;Gantait and Mukherjee 2021). Although they are not original tissues, they are able to produce various secondary metabolites (Banerjee et al. 2012;Ludwig-Muller et al. 2014). Also, for medicinal plants, hairy roots provide an effectual way to protect their natural resources by producing invaluable active ingredients in vitro (Su et al. 2022;Gantait and Mukherjee 2021). In addition, hairy roots are genetically stable than cell suspension cultures (Ibanez et al. 2016;Nguyen et al. 2021;Shi et al. 2019). In our previous studies, we used hairy roots to identify more than ten functional genes, such as auxin-responsive protein IAA (ARPI), UDP-Galactose/Glucose-4-epimerase (UGE), β-amyrin synthase (β-AS), chalcone synthase (CHS), and chalcone isomerase (CHI) genes, involved in the secondary metabolism of triterpenoids and flavonoids in G. uralensis (Hou et al. 2021;Wang et al. 2021;Yin et al. 2020). Therefore, hairy root culturing is an effective method to study gene expression and function in plants.
In summary, CRTZ, COMT, and F3′H are key enzymes involved in the synthesis of tetraterpenoids, lignins, and flavonoids, which compete with the production of GA for substrates from primary metabolism (Fig. 1). In our previous transcriptome study, we found that the expression levels of CRTZ, COMT, and F3′H correlate inversely with GA contents in G. uralensis . Based on this observation, we hypothesize that CRTZ, COMT, and F3′H negatively regulate GA synthesis. This study is to test the hypothesis and determine the roles of these genes on GA production. Accordingly, we performed gene editing and overexpression analyses of these three genes in hairy roots of G. uralensis. The findings of this work may provide new insights on the functions of CRTZ, COMT, and F3′H in the dynamic biosynthetic network of GA in G. uralensis.

Plant materials
Plump and healthy G. uralensis seeds were selected and surface-sterilized with 1% mercury bichloride and 75% ethanol solutions, and then planted on Murashige and Skoog (MS) medium for about 2 weeks. The seedlings were identified based on the internal transcribed spacer (ITS) sequences as described in our previous study ).

Target genes cloning
Total RNA was extracted from the above aseptic seedlings using an All-In-One DNA/RNA Mini-Preps Kit (Sangon Biotech (Shanghai) Co., Ltd., China) following the operating instruction. cDNA was converted using a First Strand cDNA Synthesis Kit RT0212-01 (Biomiga. Inc., China). Primer pairs used for amplifying CRTZ, COMT, or F3′H were designed by Primer-BLAST and listed in

Construction of plant binary expression vectors and CRISPR/Cas9 vectors
We constructed plant binary vectors expressing CRTZ, COMT, or F3′H as shown in Fig. S1a. The specific primers used for amplifying target genes are listed in s), and 72 °C 5 min. Target genes were inserted into linearized pCAMBIA1305.1 between the restriction enzyme cutting sites Spe I and Bgl II using a BM effusion kit (Beijing Biomed Medical Technology Co., Ltd., Beijing, China) at 50 ℃ for 30 min to yield recombinant plasmids, pCA-COMT, pCA-CRTZ, and pCA-F3′H. We also constructed plant CRISPR/Cas9 vectors for knocking out CRTZ, COMT, or F3′H as shown in Fig. S1b. The sgRNA sequences were designed and listed in Table S3. They were inserted into linearized pHSE401 between the two Bsa I sites by T4 ligase at room temperature for 10 min to construct the CRISPR/Cas9 plasmids, pHSE-COMT, pHSE-CRTZ, and pHSE-F3′H. The above recombinant plasmids were all electrotransfered into Agrobacterium rhizogenes ATCC15834 using a Gene Pulser Xcell™ electroporation instrument (Bio-Rad, USA) (PC: 200 Ω; C: 25 μF; U: 2400 V). Also, the empty pCAMBIA1305.1 and pHSE401 vectors without exogenous genes were introduced into A. rhizogenes ATCC15834 at the same electroporation conditions. Next, the positive colonies were selected on Tryptone-Yeast extract (TY) plates containing kanamycin (50 mg•L −1 ) at 28 ℃. PCR and sequencing were performed to identify the correct vectors and strains.

Establishment of G. uralensis hairy root lines
The establishment of hairy root lines was performed as described in our previous study . The above recombinant A. rhizogenes were cultured in liquid TY medium containing 50 mg•L −1 kanamycin at 28 ℃, 200 rpm, and collected in the logarithmic growth phase, then resuspended with the same volume of liquid 6,7-V medium. The G. uralensis aseptic seedlings were cut and the hypocotyls and cotyledons were used as explants. They were wounded by scalpel and soaked in the above resuspended cultures for 25 min at room temperature to complete the infection. Next, the infected explants were cultured on 6,7-V plates at 25 °C and dark for 2 days, and then transferred onto 6,7-V plates containing cefotaxime sodium (Cef) (500 mg·L −1 ) and cultured at 25 °C and dark. When hairy roots were ~ 5 cm, they were cut from explants and trans-cultured on fresh 6,7-V plates containing Cef at 25 °C and dark. In total, we induced nine groups of G. uralensis hairy roots using different recombinant A. rhizogenes strains. The wild type (WT) was induced by normal A. rhizogenes ATCC15834. The negative control hairy root lines were induced by A. rhizogenes strains containing empty pHSE401 or pCAM-BIA1305.1, referred as NC-PHSE or NC-PCA, respectively. The hairy root lines overexpressing COMT, CRTZ, or F3′H, referred as COMT + , CRTZ + , or F3′H + , were induced by recombinant A. rhizogenes strains containing pCA-COMT, pCA-CRTZ, or pCA-F3′H, respectively. The hairy root lines knocking out COMT, CRTZ, or F3′H, referred as COMT − , CRTZ − , or F3′H − , were induced by recombinant A. rhizogenes strains containing pHSE-COMT, pHSE-CRTZ, or pHSE-F3′H, respectively. Hairy roots were cultured on 6,7-V plates (25 °C, dark), and gradient concentration of Cef (500, 300, and 100 mg·L −1 ) was used to eliminate the residual A. rhizogenes until they were cleared.

Identification of G. uralensis hairy root lines
Genomic DNA of G. uralensis hairy root lines were extracted by All-In-One DNA/RNA Mini-Preps Kit (Sangon Biotech (Shanghai) Co., Ltd., China) and used as PCR templates. rolC gene is a specific marker gene in hairy roots. We amplified rolC genes in all the G. uralensis hairy root lines using a forward primer of 5′-CAT ATA TGC CAA ATT TAC ACTAG-3′ and a reverse primer of 5′-GTT AAC AAA CTA GGA AAC AGG-3′, at PCR cycling parameters of 95 °C for 5 min, 34 cycles of 95 °C for 30 s, 56 °C for 30 s, 72 °C for 60 s, and 72 °C for 5 min. For identification of the COMT + , CRTZ + , and F3′H + hairy root lines, we respectively amplified the corresponding target genes using the PCR primers listed in Table S1 and the programs described in Section "Target genes cloning". To analyze the gene editing sites in the COMT − , CRTZ − , and F3′H − hairy root lines, we designed specific primers (Table S4) to respectively amplify the corresponding exons of COMT, CRTZ, and F3′H. The PCR cycling parameters were as follows: 95 °C for 5 min, 32 cycles of 95 °C for 30 s, 55.5 °C for 50 s, 72 °C for 60 s, and 72 °C for 5 min. PCR products were cloned into pMD™ 19-T Cloning Vector (Takara Biomedical Technology (Beijing) Co., Ltd., China) and selected on LB plates supplemented with ampicillin (50 mg·L −1 ) at 37 °C. Single colonies were randomly picked for sequencing analysis which was performed by Sangon Biotech (Shanghai) Co., Ltd., China.

Gene expression levels analysis
Real-time quantitative PCR (RT-qPCR) was carried out with β-actin gene as the internal control to analyze the expression levels of COMT, CRTZ, or F3′H in hairy root lines overexpressing the three genes. Primers for RT-qPCR are listed in Table S5. Total RNA of the COMT + , CRTZ + , and F3′H + lines were extracted using All-In-One DNA/RNA Mini-Preps Kit (Sangon Biotech (Shanghai) Co., Ltd., China) and reversely transcribed into cDNA by First Strand cDNA Synthesis Kit RT0212-01 (Biomiga. Inc., China). RT-qPCR was carried out using an Agilent StrataGene Mx3005P QPCR (Agilent Technologies, USA) with a program as follows: 95 ℃ 3 min, 45 cycles (95 ℃ 7 s, 57 ℃ 10 s, 72 ℃ 15 s). The expression level differences were evaluated by the 2 −∆∆Ct method in triplicate (Annaratone et al. 2013).

Suspension culture of hairy roots and samples preparation
The validated hairy roots were isolated from each line and trans-cultured in liquid 6,7-V medium (250 mL-conical flasks, 110 rpm, 25 °C, dark). After hairy root lines adapted to the liquid culture system and thrived, the healthy branches (2.0 g) were cut in triplicate and suspension cultured in fresh 6,7-V medium at the same conditions for three weeks. Next, the collected hairy roots were dried at 65 °C to constant weight and powdered for content analyses of GA, lignin, and total flavonoids.

GA content assay by UPLC
The UPLC method for determination of the GA contents in G. uralensis hairy roots established in our previous study was adopted . The standard compound GA with a purity of 99.45% was purchased from China National Institutes for Food and Drug Control. The stock solution of GA was prepared with 50% methanol at a concentration of 0.0904 mg‧mL −1 . The linear curve was evaluated at seven points by gradient dilution of the GA stock solution with 50% methanol to 0.09040, 0.07232, 0.05424, 0.04520, 0.01808, 0.00904, and 0.00452 mg‧mL −1 . 100 mg of each hairy root powder sample was extracted with 50 mL 50% methanol using ultrasonic for 30 min (40 kHz, 500 W), and filtered with 0.45 μm filter membranes. The chromatographic conditions were as follows: chromatographic instrument: Waters Acquity UPLC system (Waters Corporation, Milford, Massachusetts, USA), chromatographic column: Waters UPLC BEH C 18 column (2.1 mm × 100 mm, 1.7 μm), UV detection wavelength: 250 nm, column temperature: 40 °C, flow rate: 0.3 mL·min −1 , and injection volume: 1 μL. The mobile phase A was acetonitrile and B was 0.05% phosphoric acid. The UPLC gradient elution program was shown in Table S6.

Lignin content assay by UV spectrophotometry
To evaluate the effects of COMT overexpression or knockout on the lignin metabolism, we used an UV spectrophotometry method (Xu et al. 2022) to detect the lignin contents in the WT, NC-PCA, NC-PHSE, COMT + , and COMT − hairy root lines. The lignin contents were measured in triplicate using the lignin assay kit (Beijing Solarbio Science & Technology Co., Ltd, Beijing, China) at 280 nm on an Epoch microplate reader (Bio Tek Instruments Inc, USA).

Total flavonoids determination by UV spectrophotometry
To evaluate the effects of F3′H over-expression or knockout on the flavonoid metabolism, we detected the total flavonoid contents in the WT, NC-PCA, NC-PHSE, F3′H + , and F3′H − hairy root lines according to a previous study (Dang et al. 2021). The standard compound liquiritin with a purity of 98.50% was purchased from China National Institutes for Food and Drug Control. The stock solution of liquiritin was prepared with methanol at a concentration of 0.152 mg‧mL −1 . The linear curve was evaluated at six points by gradient dilution of the liquiritin stock solution with methanol to 0.0019, 0.0038, 0.0076, 0.0152, 0.0304, and 0.0608 mg‧mL −1 . 20 mg of each hairy root powder sample was extracted with 10 mL methanol using ultrasonic for 80 min (20 kHz, 250 W) and filtered. 0.5 mL of the subsequent filtrate was transferred into a 10 mL volumetric flask, then 1 mL methanol and 0.5 mL 10% KOH were incorporated into the flask for chromogenic reaction at room temperature for 5 min. After the reaction, methanol was added into the flask to a final volume of 10 mL and the absorbance at 334 nm was measured on an Epoch microplate reader (Bio Tek Instruments Inc., USA).

Statistical and bioinformatics analysis
The IBM SPSS Statistics 20 was used to perform statistical analyses. BLAST (http:// www. ncbi. nlm. nih. gov) and DNA-MAN 6.0.3.99 were used to analyze the sequencing results. MEGA 11 was applied to construct the phylogenetic tree using the Maximum Composite Likelihood method.

Identification of three target genes
As shown in Fig. S2a, three fragments, the length of which were about 1200, 1000, and 1000 bp, were obtained by PCR. Sequencing results confirmed that the exact full length of the three PCR products were 1140, 930, and 993 bp, respectively. The 1140 bp-PCR product had 86.71% identity with the Abrus precatorius COMT (GenBank accession No. XM_027512828.1), the 930 bp-PCR product had 89.82% identity with the Glycine max CRTZ (NM_001254504.1), and the 993 bp-PCR product had 81.82% identity with Cajanus cajan F3′H (XM_020357268.2), indicating that the COMT, CRTZ, and F3′H sequences were obtained from G. uralensis. We registered these sequences in GenBank with the accession numbers as follows: COMT (MZ169549), CRTZ (MZ169550), and F3′H (MZ169551). It is the first time that the COMT, CRTZ, and F3′H genes are identified from G. uralensis. To evaluate the evolution of these genes, we then established phylogenetic trees based on COMT, CRTZ, and F3′H homologous sequences registered in NCBI, respectively. As shown in Fig. 2, COMT and F3′H homologous sequences both clustered into three major independent clades, Dicotyledoneae, Monocotyledoneae, and Pteridophyta, while CRTZ orthologs clustered into four, Dicotyledoneae, Monocotyledoneae, Pteridophyta, and Bryophyta. These findings suggest that the evolution of the three enzymes is conserved.

Construction of recombinant vectors
As shown in Fig. S2b, three fragments were amplified from the plant binary expression vectors, pCA-COMT, pCA-CRTZ, and pCA-F3′H, respectively. Sequencing results further confirmed that these PCR products were completely identical to COMT (MZ169549), CRTZ (MZ169550), and F3′H (MZ169551) cloned from G. uralensis. These findings demonstrated that the plant binary expression vectors for overexpressing COMT, CRTZ, or F3′H were correct. As shown in Fig. S2c, several 400 bp PCR products were obtained from the plant CRISPR/Cas9 vectors, pHSE-COMT, pHSE-CRTZ, and pHSE-F3′H. Sequencing results confirmed that these fragments contained the corresponding sgRNA sequences of COMT, CRTZ, or F3′H genes, indicating that the CRISPR/Cas9 vectors for knocking out these genes were constructed successfully.

Identification of G. uralensis hairy root lines
We generated nine groups of G. uralensis hairy root lines, including WT, NC-PCA, NC-PHSE, COMT + , CRTZ + , F3′H + , COMT − , CRTZ − , and F3′H − . Figure 3 shows the growth condition of these hairy roots after the onset of induction for 10, 20, and 30 days. The roots of the transgenic lines grew at a rate similar to that of WT. To inspect these hairy root lines, we cloned rolC genes, the signature gene in hairy roots, from all lines and got 600 bp-fragments by PCR (Fig. S3a), which were identified to have 100% identity with rolC (GenBank accession No. DQ160187.1). Also, the fragments amplified from the COMT + , CRTZ + , and F3′H + lines shown in Fig. S3b were confirmed with sequence analyses. Fig. S3c and S3d display the third exons of COMT and CRTZ amplified from the COMT − and CRTZ − lines respectively, and Fig. S3e shows the first exons of F3′H amplified from the F3′H − lines. These fragments, as described in next section, will be used to analyze the exact editing sites of target genes in hairy root lines of COMT − , CRTZ − , and F3′H − through further cloning and sequencing. In the end, one WT, five COMT + , five CRTZ + , seven F3′H + , three COMT − , five CRTZ − , four F3′H − , one NC-PCA, and one NC-PHSE hairy root lines were obtained.

Analyses of gene editing and expression levels in transgenic G. uralensis hairy root lines
Further cloning and sequencing analyses confirmed that the COMT gene was edited in four hairy root lines (COMT − -2, -5, -6, and -8) out of nine with a gene editing efficiency of 44.4%. The gene editing details are shown in Fig. 4a and the mutation sites present in COMT amino acid sequences are listed in Table 1. We observed a synonymous mutation in the COMT − -2 line and missense mutations in the other three lines. The most efficient gene edition was a fragment deletion in the COMT − -5 line, which results in a termination codon "TAA" in this position.
CRTZ gene was edited in eight lines (CRTZ − -1, -2, -3, -4, -5, -6, 7, and -8) out of ten with an editing efficiency of 80%. Among them, homozygous mutations were present in four lines, CRTZ − -1, -3, -4 and -6 ( Fig. 4b), while heterozygous mutations in the other four (Fig. S4b). Frameshifts were observed in all lines. The mutation sites in the CRTZ amino acid sequences are listed in Table 2. The CRTZ sequences in the CRTZ − -1 and CRTZ − -6 lines are the same, both containing two mutation sites. The CRTZ sequence in the CRTZ − -3 line is identical to that in the CRTZ − -4 line with a single mutation site. All mutation sites in these four lines were located within the functional domain of CRTZ.
We next examined the gene expression levels of the three genes in newly constructed COMT + , CRTZ + , and F3′H + hairy root lines. As illustrated in Fig. 4d, the relative expression levels of COMT, CRTZ, and F3′H in the COMT + , CRTZ + , and F3′H + hairy root lines were all remarkably higher than that in WT. In particular, samples COMT + -4, CRTZ + -4, and F3′H + -6 showed the highest expression level in their respective groups. Figure 5a shows the hairy root samples cultured in liquid 6,7-V medium. Figure 5b shows the collected hairy root samples, which were prepared for UPLC analyses. It is worth noting that the roots of the F3′H − and CRTZ + lines had a whitish appearance and those of the CRTZ − lines appeared reddish. The hairy roots of other lines were whitish with a yellow undertone. The UPLC chromatograms of GA in the WT, NC-PCA, NC-PHSE, COMT + , CRTZ + , F3′H + , COMT − , CRTZ − , and F3′H − lines as well as the GA standard are shown in Fig. 6a. The UPLC retention time of GA was 6.532 min and the standard curve was expressed as: Y = 2,683,455.37 X − 1660.17 (R 2 = 0.9999, X: the GA content (mg·mL −1 ), Y: the peak area (mAU * s)). The GA content in all of the hairy root lines is calculated and listed in Table S7. The differences of GA content among different lines are shown in Fig. 6b and those among different groups are shown in Fig. 6c. We found that the GA content in negative control (NC-PCA and NC-PHSE) was comparable to that in WT. Interestingly, we noticed an inverse correlation between the expression levels of the three genes and the GA Fig. 3 The induction and culture of G. uralensis hairy roots. The photos of 10-, 20-, and 30-day old hairy root lines after the onset of induction in solid 6,7-V medium. WT represents the wild type hairy roots. NC-PCA and NC-PHSE represent the negative control hairy root lines containing the empty pCAMBIA1305.1 and pHSE401 vec-tors, respectively. F3′H − , COMT − , and CRTZ − represent the hairy root lines knocking out the F3′H, COMT, and CRTZ genes, respectively. F3′H + , COMT + , and CRTZ + represent the hairy root lines overexpressing the F3′H, COMT, and CRTZ genes, respectively 1 3 Fig. 4 Analyses of gene editing sites and expression levels in hairy root lines. a-c show the homozygous mutations in the COMT − , CRTZ − , and F3′H − hairy root lines, respectively. The mutation sites are marked by the red triangles and the corresponding sites in the wild type sequences are marked by the blue ones. d shows the relative expression levels of the three genes in the F3′H + , COMT + , and CRTZ + hairy root lines. * represents P < 0.05 (vs. WT). (Color figure online) content in hairy root lines. For example, the GA content in all the seven F3′H + lines was significantly lower than that in the WT and NC-PCA lines, while the GA content in the F3′H − lines (F3′H − -2, -8, and -10), except F3′H − -7, was significantly higher than that in the WT and NC-PHSE lines. Similar variations in GA content were also observed in the COMT + , CRTZ + , COMT − , and CRTZ − groups. As shown in Table S7, the GA content in the F3′H + (average value: 2.3654 mg·g −1 dry weight (DW)), CRTZ + (2.2050 mg·g −1 DW), and COMT + (2.3509 mg·g −1 DW) groups was significantly lower than that in both WT (3.3242 mg·g −1 DW) and NC-PCA (3.5188 mg·g −1 DW). While the GA content in the F3′H − (5.4744 mg·g −1 DW), CRTZ − (5.3651 mg·g −1 DW) and COMT − (4.2746 mg·g −1 DW) groups was significantly higher than that in WT and NC-PHSE (3.3459 mg·g −1 DW). These findings suggest that the COMT, CRTZ, and F3′H genes negatively regulate GA production.

The lignin and total flavonoids contents analyses
The lignin contents in the hairy roots of the WT, negative controls, COMT + , and COMT − lines were examined using a lignin assay kit. As shown in Fig. 7a and Table S8, we found that the lignin content in all the three COMT − lines was significantly lower than that in the WT and NC-PHSE lines, whereas the lignin content in all the five COMT + lines was significantly higher than that in the WT and NC-PCA lines. The lignin content in the COMT − group (average value: 37.23 mg·g −1 DW) was remarkably lower than that in the COMT + group (72.23 mg·g −1 DW). These findings suggest that COMT positively regulates the biosynthesis of lignin. Figure 7b and Table S9 show the total flavonoid contents in the hairy roots of the WT, negative controls, F3′H + , and F3′H − lines. We constructed a standard curve expressed as: Y = 3.9146 X − 0.0003 (R 2 = 0.9979) using X axis for the total flavonoid contents (mg·mL −1 ) and Y axis for the relative absorbance value. We found that the total flavonoid content in three F3′H − lines (F3′H − -2, -8, and -10) was significantly lower than that in the WT and NC-PHSE lines, whereas the flavonoid content in all the seven F3′H + lines was significantly higher than that in the WT and NC-PCA lines. The total flavonoid content in the F3′H − group (average value: 31.70 mg·g −1 DW) was significantly lower than that in the F3′H + group (51.00 mg·g −1 DW). These findings indicate that overexpression of F3′H promotes the biosynthesis of

Discussion and conclusion
In the present study, we established transgenic G. uralensis hairy root lines to investigate the effects of the CRTZ, COMT, and F3′H genes on GA accumulation. Through UPLC analyses, we found that in comparison with WT and negative control, the GA content was significantly lower in the hairy roots overexpressing any of the three genes. However, the content was higher in the hairy roots knocked out any of the genes. These findings suggest that the CRTZ, COMT, and F3′H genes negatively regulate GA biosynthesis. CRTZ, COMT, and F3′H are key enzymes, respectively, involved in the synthesis of tetraterpenoids, lignin, and flavonoids in G. uralensis. These pathways and the triterpenoid biosynthetic pathway, which produces GA, compete with each other for primary metabolites and hence coordinate the metabolic balance in G. uralensis. Several lines of evidence have demonstrated that overexpression of CRTZ increases the utilization of β-carotene (Pollmann et al. 2017), while knockout of CRTZ has an opposite effect (Tomlekova et al. 2021). It is thus expected that an accumulation of β-carotene induced by the CRTZ knockout leads to a feedback inhibition of the MEP pathway (Mitra et al. 2021;Tian et al. 2014), which in turn promotes the MVA pathway and the production of GA in G. uralensis. This notion is supported by the color changes of the hairy roots of the CRTZ − and CRTZ + lines. We find that the color of the CRTZ − hairy roots looks reddish, indicating a high content of β-carotene, whereas the color of the CRTZ + hairy roots appears pale, suggesting a low content of β-carotene. It has also been observed in Brassica rapa and Oncidium hybridum that down-regulation of CRTZ causes the accumulation of β-carotene and makes the flower color to turn from yellow to orange (Chiou et al. 2010;Zhang et al. 2019). It is thus likely that knocking out CRTZ inhibits the MEP pathway, which shunts more DMAPP and IPP into the MVA pathway, eventually leading to an increase of GA content in G. uralensis hairy roots.
We analyzed both contents of lignin and GA in hairy roots overexpressing or knocking out COMT to identify the function of this gene in G. uralensis. We found that overexpression of COMT promoted the production of lignin but reduced the accumulation of GA, whereas knockout of COMT had an opposite effect. Since the synthesis of GA and lignin both consumes substrates from glycolysis, the increase of lignin is bound to suppress the carbon flux into the MVA pathway, which may count for the negative effect of COMT on GA accumulation. In accordance with our findings, in Arabidopsis thaliana, knocking out COMT also caused the decrease of lignin and the increase of anthocyanin, suggesting a competitive inhibition between different secondary metabolic pathways (Xie et al. 2019;Li et al. 2020). Similarly, we observed that over-expression of F3′H promoted the accumulation of total flavonoids while inhibited the production of GA, in contrast to what we observed in hairy roots knocking out F3′H. In A. thaliana, Nicotiana tabacum, Fagopyrum tataricum, and Camellia nitidissima, over-expression of F3′H also increased the contents of total flavonoids and various polyphenol compounds, indicating a positive regulatory effect of F3′H on flavonoid production Jiang et al. 2021). Our previous study demonstrated that the promotion of flavonoid synthesis reduced GA accumulation in G. uralensis . Therefore, over-expression of F3′H promotes flavonoid production at the expense of triterpenoid synthesis, which may count for the negative effect of F3′H on GA production.
Although the GA content differs greatly between different groups, within the same group, the content in different lines also varies. This phenomenon is mainly observed in groups knocking out the three genes. For example, in the F3′H − group the GA content in three lines (F3′H − -2, -8, and -10) is significantly higher than that in WT and NC-PHSE. However, the F3′H − -7 line is an exception. We believe that the variation in the GA content among the F3′H − lines is caused by the different gene editing effects. As shown in Table 3, there is only one mutation at the Fig. 5 UPLC sample preparation. a shows the 3-week old hairy root lines cultured in liquid 6,7-V medium. b shows the hairy root samples collected for UPLC analyses 56th position causing a substitution of valine to isoleucine in the F3′H sequence of F3′H − -7 line, while multiple sites are mutated in the other three lines, which may exacerbate the dysfunction of F3′H. This notion is verified by the total flavonoid content analysis. We observed that the total flavonoid content in the F3′H − -2, -8, and -10 lines was significantly decreased, but that in the F3′H − -7 line was unaffected (Fig. 7b). Also, in the COMT − group the GA content of the COMT − -5 line was significantly higher than that in WT and NC-PHSE. In the other two lines (COMT − -6 and 8), although the GA content was also higher, the differences were not remarkable. We analyzed Fig. 6 GA content analyses in hairy root lines. a shows the UPLC chromatograms, line 1 is the UPLC chromatogram of the reference substance GA, lines 2-10 are UPLC chromatograms of the WT, NC-PCA, NC-PHSE, F3′H + , COMT + , CRTZ + , F3′H − , COMT − , and CRTZ − samples, respectively. b shows the GA content in each hairy root line. c shows the GA contents in the nine groups. * represents P < 0.05 (vs. WT). # represents P < 0.05 (vs. NC-PCA or NC-PHSE) the gene editing in these three lines and found that the mutation in COMT cDNA sequence of the COMT − -5 line created a premature termination codon, which makes the translation of COMT stop at the 160th position, leading to a complete loss of COMT function ( Fig. 4a; Table 1). In the COMT sequence of the COMT − -6 line, there is only one mutation at the 174th position causing a substitution of methionine to histidine and in that of the COMT − -8 line only one mutation at the 219th position causing a substitution of valine to alanine. These mutations are expected to affect the function of COMT, but not as serious as the mutation in the COMT − -5 line that causes a premature termination of the COMT translation. This notion is supported by the lignin content analysis of the COMT − lines. As shown in Fig. 7a, although the lignin content in the three COMT − lines is reduced, that in the COMT − -5 line is the lowest. Also, there is a significant difference in lignin content between the COMT − -5 line and the other two lines (p < 0.001). In addition, the GA content differences among the CRTZ − lines might also due to the variations in gene editing effects ( Fig. 4b; Table 2).
To sum up, this is the first time that CRTZ, COMT, and F3′H genes are identified from G. uralensis and their functions are investigated through gene knockout and overexpression hairy root models. Our findings indicate a negative regulatory effect of CRTZ, COMT, and F3′H on GA production due to the competition among different secondary metabolic pathways. In future studies, we will evaluate functions of many other key genes and establish a genic regulatory network responsible for GA synthesis in G. uralensis.

3
Author contributions YL conceived and designed the experiments. ZXZ, WWD and ZYC performed the experiments. WPX and DDW helped in analyzing the relevant data and searching literature. YL and TGL wrote the manuscript.

Funding
The authors have not disclosed any funding.

Declarations
Conflict of interest The authors declare that there is no conflict of interest regarding the publication of this paper.