Histone Deacetylase GiSRT2 Negatively Regulates Flavonoid Biosynthesis in Glycyrrhiza inflata

Glycyrrhiza inflata Batalin is a medicinal licorice species that has been widely used by humans for centuries. Licochalcone A (LCA) is a characteristic flavonoid that accumulates in G. inflata roots with high economical value. However, the biosynthetic pathway and regulatory network of its accumulation remain largely unknown. Here we found that a histone deacetylase (HDAC) inhibitor nicotinamide (NIC) could enhance the accumulation of LCA and total flavonoids in G. inflata seedlings. GiSRT2, a NIC-targeted HDAC was functionally analyzed and its RNAi transgenic hairy roots accumulated much more LCA and total flavonoids than its OE lines and the controls, indicating a negative regulatory role of GiSRT2 in the accumulation of LCA and total flavonoids. Co-analysis of transcriptome and metabolome of RNAi-GiSRT2 lines revealed potential mechanisms in this process. An O-methyltransferase gene, GiLMT1 was up-regulated in RNAi-GiSRT2 lines and the encoded enzyme catalyzed an intermediate step in LCA biosynthesis pathway. Transgenic hairy roots of GiLMT1 proved that GiLMT1 is required for LCA accumulation. Together, this work highlights the critical role of GiSRT2 in the regulation of flavonoid biosynthesis and identifies GiLMT1 as a candidate gene for the biosynthesis of LCA with synthetic biology approaches.


Introduction
Licorice is a popular medicinal herb that has been used by humans worldwide for centuries. The main active components of licorice are triterpenoids and flavonoids. Among them, licochalcone A (LCA) is the characteristic flavonoid in Glycyrrhiza inflata Batalin [1], one of the major medicinal licorice species. LCA has a wide range of pharmacological properties. To date, it has been discovered that LCA has excellent activity to inhibit the proliferation and motility of 17 classes of cancer cells [2,3]. Furthermore, LCA has demonstrated anti-inflammatory effects against both acute and chronic inflammation [4,5]. Moreover, LCA has been shown to have antioxidant activity and neuroprotective effects [6,7]. LCA is also a valuable compound used in the pharmaceutical and cosmetics industries. However, the limited yield of LCA restricts its application, which motivates research into methods for improving LCA production.
Despite its great commercial value, LCA only accumulates in low abundance in a very limited number of plant species. G. inflata is the major resource for LCA on the market.

RNA Extraction and qRT-PCR Analysis
RNA extraction and qRT-PCR were carried out as previously described [20]. GiCOPS3 was used as internal reference gene [21]. All primers used in qRT-PCR are listed in Table S1.

Vector Construction
The genome sequence of G. inflata used in this study was provided by our research group (https://ngdc.cncb.ac.cn/; NGDC; CRA009044; accessed on 25 November 2022). The coding sequences (CDS) of GiSRT2 (accession number OQ982386) and GiLMT1 (accession number OQ982385) were PCR-amplified from cDNAs, digested with Sal I and Kpn I, or Pst I and Spe I enzymes, respectively, and cloned into pSuper1300-GFP to generate OE-GiSRT2 and OE-GiLMT1 transgenic hairy roots, respectively. The pRNAiGG vector [22] was used to generate RNAi-GiSRT2 and RNAi-GiLMT1 transgenic hairy roots. The genespecific fragments were PCR amplified and cloned into the pRNAiGG vector using the Bsa I restriction site. The CDS of GiLMT1 was cloned into the E. coli expression vector pColdII (TaKaRa) using Sac I and Sal I restriction sites to produce a recombinant protein. The full-length CDS of GiSRT2 was cloned into pCAMBIA1300-UBQ-GFP in frame with GFP for subcellular localization analysis. All vectors were confirmed using DNA sequencing (Qingke, Beijing, China).

Generation of Transgenic Hairy Roots of G. inflata
Transgenic hairy roots were generated as previously described with minor modifications [11]. Binary vectors were introduced into the modified A. rhizogenes strain MSU440. Three different types of controls were used in this work, including non-transgenic hairy roots (WT), OE empty vector-induced hairy roots (OE-CK), and RNAi empty vector-induced hairy roots (RNAi-CK). PCR with rolB (from Ri plasmid of MSU440) and gene-specific primers was performed to identify the positive transgenic lines. All the primers used in this study are listed in Table S1.

Compound Extraction and Determination
To extract the metabolites, 10 mg of freeze-dried samples was subjected to ultrasonicassisted extraction twice with 1 mL of methanol for 30 min. After centrifugation at 12,000× g for 10 min, the supernatants were dried and finally redissolved with 300 µL of methanol.
Total flavonoids content was measured using the sodium nitrite-aluminum nitrate colorimetric method using rutin as the standard [23].

Transcriptomic and Metabolomic Analysis
Hairy roots of RNAi-GiSRT2 and the corresponding CK lines (RNAi empty vectorinduced hairy roots) were sampled for RNA-seq and metabolite analyses by Metware Biotechnology Co., Ltd. (Wuhan, China) as previously described [24]. Differentially expressed genes (DEGs) were selected with |log 2 (Fold Change)| ≥ 1 (|log 2 FC| ≥ 1) and p-value < 0.05. Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses of DEGs were implemented using the OmicShare tools accessed on 9 February 2023 (http://www.omicshare.com/tools). The differentially accumulated metabolites (DAMs) were screened according to the variable importance of the projection (VIP ≥ 1) produced by orthogonal projections to latent structures-discriminant analysis (OPLS-DA) and univariate analysis of variance (ANOVA, p < 0.05).

Enzyme Assay of GiLMT1
A single colony of E. coli BL21 (DE3) carrying the GiLMT1/pColdII vector was incubated in 0.2 mL of Luria-Bertani (LB) medium containing 50 µg/mL ampicillin at 37 • C with agitation at 200 rpm for 16 h. The inoculum was then added into 5 mL medium and cultured until the OD 600 reached 0.6, followed by the addition of isopropyl-β-Dthiogalactopyranoside (IPTG) to a final concentration of 0.1 mM. The culture was then incubated at 16 • C with shaking at 140 rpm for 16 h. Cells were collected using centrifugation (4000× g, 10 min) and resuspended in 5 mL Tris-HCl (50 mM, pH 7.5). The cells were disrupted using an ultrasonic homogenizer (SCIENTZ-IID) at 50 W for 10 min, and the supernatants were used for enzyme assays.

Subcellular Localization Analysis of GiSRT2
The subcellular localization of GiSRT2 was studied using transient expression in Arabidopsis protoplasts as previously reported [20].

Effects of NIC Treatment on Seedling Growth and LCA Accumulation in G. inflata
To test the effects of NIC on seedling growth and LCA accumulation, 7-day-old G. inflata seedlings were treated with 1 mM NIC on 1/2MS plates. After 3, 5, and 7 days of treatment, NIC significantly inhibited root growth, but resulted in a higher level of LCA than the mock control ( Figure 1A-C). Total flavonoids content was also measured 7 days after the treatment. Similar with the result obtained for the characteristic flavonoid LCA, 1 mM NIC promoted the accumulation of total flavonoids, indicating a systemic activation of flavonoid biosynthesis in G. inflata roots ( Figure 1D). Since NIC is a specific inhibitor of SIR class HDACs, this result suggested that SIR class HDACs are required for normal growth, and they may play negative roles in the regulation of the accumulation of flavonoids, especially LCA.

GiSRT2 Negatively Regulates LCA Accumulation
There are two SIR class HDACs predicted in the G. inflata genome named GiSRT1 and GiSRT2. GiSRT1 was neither detected in 1, 2, or 3-year-old roots nor leaves in the qRT-PCR assay. Whereas, GiSRT2 was detectable and the expression levels were higher in leaves than in roots ( Figure S2A). Since LCA is accumulated in G. inflata roots [25] but not detectable in shoots, it suggests that GiSRT2 is likely the negative regulator of LCA accumulation. We then studied the subcellular localization of GiSRT2-GFP in Arabidopsis protoplast cells. The confocal images showed that the GFP signal of GiSRT2-GFP overlapped with the mCherry signal of NLS-mCherry, indicating the nucleus localization of GiSRT2-GFP ( Figure S2B).

GiSRT2 Negatively Regulates LCA Accumulation
There are two SIR class HDACs predicted in the G. inflata genome named and GiSRT2. GiSRT1 was neither detected in 1, 2, or 3−year−old roots nor leav qRT−PCR assay. Whereas, GiSRT2 was detectable and the expression levels we in leaves than in roots ( Figure S2A). Since LCA is accumulated in G. inflata roots not detectable in shoots, it suggests that GiSRT2 is likely the negative regulato accumulation. We then studied the subcellular localization of GiSRT2−GFP in Ara protoplast cells. The confocal images showed that the GFP signal of GiSRT2−G lapped with the mCherry signal of NLS−mCherry, indicating the nucleus locali GiSRT2−GFP ( Figure S2B).
To verify the function of GiSRT2 in the regulation of LCA accumulation, w ated OE−GiSRT2 and RNAi−GiSRT2 transgenic hairy roots of G. inflata. The tr hairy root lines were confirmed using PCR and qRT−PCR assays (Figures S3 an Since the growth of hairy roots or accumulation of specialized metabolites am (generated by infection with MSU440), OE−CK (generated by infection with MSU rying pSuper−GFP empty vector), and RNAi−CK (generated by infection with carrying pRNAiGG empty vector) ( Figure S1) were similar, we used WT transge roots as the control in this part. The RNAi−GiSRT2 hairy roots exhibited a dar compared to the control and OE−GiSRT2 lines, consistent with the higher leve flavonoids in RNAi−GiSRT2 lines (Figure 2A,D). Furthermore, the RNAi−GiSR genic hairy roots accumulated a higher level of LCA when compared with the W lines and OE−GiSRT2 lines. Based on the reported data, we proposed a hypothet To verify the function of GiSRT2 in the regulation of LCA accumulation, we generated OE-GiSRT2 and RNAi-GiSRT2 transgenic hairy roots of G. inflata. The transgenic hairy root lines were confirmed using PCR and qRT-PCR assays (Figures S3 and 2B,C). Since the growth of hairy roots or accumulation of specialized metabolites among WT (generated by infection with MSU440), OE-CK (generated by infection with MSU440 carrying pSuper-GFP empty vector), and RNAi-CK (generated by infection with MSU440 carrying pRNAiGG empty vector) ( Figure S1) were similar, we used WT transgenic hairy roots as the control in this part. The RNAi-GiSRT2 hairy roots exhibited a darker color compared to the control and OE-GiSRT2 lines, consistent with the higher level of total flavonoids in RNAi-GiSRT2 lines (Figure 2A,D). Furthermore, the RNAi-GiSRT2 transgenic hairy roots accumulated a higher level of LCA when compared with the WT control lines and OE-GiSRT2 lines. Based on the reported data, we proposed a hypothetical LCA biosynthesis pathway ( Figure S4) [9,10,25]. The contents of specialized flavonoids in this pathway including LCA precursor echinatin, its isomer licochalcone C (LCC), and other reported bioactive flavonoid isoliquiritigenin in licorice roots were measured. HPLC results showed that the contents of echinatin and isoliquiritigenin, as well as LCC, were much higher in RNAi-GiSRT2 lines than those in the control and OE-GiSRT2 lines ( Figure 2D-F). These results demonstrated that GiSRT2 negatively regulated the accumulation of LCA and total flavonoids in G. inflata hairy roots. pathway including LCA precursor echinatin, its isomer licochalcone C (LCC), and other reported bioactive flavonoid isoliquiritigenin in licorice roots were measured. HPLC results showed that the contents of echinatin and isoliquiritigenin, as well as LCC, were much higher in RNAi−GiSRT2 lines than those in the control and OE−GiSRT2 lines ( Figure  2D-F). These results demonstrated that GiSRT2 negatively regulated the accumulation of LCA and total flavonoids in G. inflata hairy roots.

Transcriptome Sequencing (RNA−Seq) Analysis of RNAi−GiSRT2 Lines
To unveil the mechanism underlying the enhanced LCA biosynthesis in RNAi−GiSRT2 transgenic hairy roots, the transcriptome sequencing was carried out. Principal component analysis (PCA) of the samples showed that the CK (RNAi empty vector−induced hairy roots) and RNAi−GiSRT2 sample types were grouped well and separated clearly, indicating the high quality of the generated transcriptome data ( Figure S5A). The significant DEGs were selected by setting |log2FC| ≥ 1 and P−adjust ≤ 0.001 as the thresholds. As shown in the volcano plot, a total of 4930 significant DEGs were identified in the RNAi−GiSRT2_VS_CK pair ( Figure S5B). Hierarchical cluster analysis showed that Expression level of GiSRT2 in the OE-GiSRT2 (B) and RNAi-GiSRT2 (C) lines were detected using qRT-PCR. GiCOPS3 was used as the internal control. The content of total flavonoids (D), LCA (E), echinatin (F), isoliquiritigenin (G), and licochalcone C (H) in OE-GiSRT2 and RNAi-GiSRT2 transgenic hairy roots were detected using HPLC. The non-transgenic hairy roots are set as the WT. The different lower-case letters indicate significant differences at a p value of 0.05 for the relative expression level and the contents of flavonoids among samples. Student's t-test, n ≥ 3.

Transcriptome Sequencing (RNA-Seq) Analysis of RNAi-GiSRT2 Lines
To unveil the mechanism underlying the enhanced LCA biosynthesis in RNAi-GiSRT2 transgenic hairy roots, the transcriptome sequencing was carried out. Principal component analysis (PCA) of the samples showed that the CK (RNAi empty vector-induced hairy roots) and RNAi-GiSRT2 sample types were grouped well and separated clearly, indicating the high quality of the generated transcriptome data ( Figure S5A). The significant DEGs were selected by setting |log 2 FC| ≥ 1 and p-adjust ≤ 0.001 as the thresholds. As shown in the volcano plot, a total of 4930 significant DEGs were identified in the RNAi-GiSRT2_VS_CK pair ( Figure S5B). Hierarchical cluster analysis showed that the expression patterns of most DEGs in CK and RNAi-GiSRT2 were completely opposite, and indicated good repeatability between biological replicates ( Figure S5C).
A total of five differentially expressed TFs and six differentially expressed structural genes were selected for qRT-PCR to verify the reliability of RNA sequencing. As shown in Figure 3A, the expression levels of the selected genes displayed high consistency with the RNA-seq data, and the RNA sequencing data correlated significantly with the qRT-PCR results (R 2 = 0.94, p < 0.05; Figure 3B). To elucidate the function of the DEGs, the GO and KEGG enrichment analysis were carried out. As shown in Figure S6A, the annotated DEGs were classified into 14 subclasses under molecular function, 15 subclasses under biological process, and 2 subclasses under cellular component. Among them, 'cellular process', 'metabolic process', and 'response to stimulus' were the most significantly enriched BPs. The majority of GO terms in cellular component enrichments occurred in 'cellular anatomical entity' and molecular function enrichments primarily occurred in 'binding' and 'catalytic activity'. KEGG annotation results showed that DEGs were assigned to 130 pathways, among these pathways, the enriched pathways were significantly concentrated in 'phenylpropanoid biosynthesis', 'isoflavonoid biosynthesis', and 'biosynthesis of secondary metabolites' (Figure S6B).
Our RNA-seq analysis revealed that the DEGs contained 252 transcription factorencoding genes, of which 171 showed up-regulation and 81 down-regulation ( Figure S7A). They could be divided into 48 different common families. These families contain a significant number of TFs, particularly the AP2/ERF (12.7%), bHLH (9.52%), and MYB-related (8.33%) families ( Figure S7B); the expression levels of identified TF family genes are shown by heatmap. The expression of some genes of WRKY, MYB, and AP2/ERF TFs was higher in RNAi-GiSRT2 lines than in the controls ( Figure S7C), indicating that these genes might play a regulatory role in the flavonoid metabolism.

Metabolome Analysis of RNAi-GiSRT2 Lines
To reveal the regulatory role of GiSRT2 in the whole metabolism, a widely targeted metabolomics assay was carried out using CK and RNAi-GiSRT2 hairy roots. We used an ultra-performance liquid chromatography-tandem mass spectrometry (UPLC-MS) method to identify changes in metabolite levels. PCA indicated that the metabolites of different genotypes were significantly different ( Figure S8A). A total of 1360 metabolites were detected with this approach, including 12 different types of substances, among these metabolites, with 172 phenolic acids, 351 flavonoid metabolites, 171 organic acids, 141 lipids, and 135 terpenoids ( Figure S8B). Cluster analysis was also performed with twelve samples being clearly divided into four groups, indicating significant differences in metabolites among four experiment groups ( Figure S8C).

Combined Analysis of Transcriptome and Metabolome
To investigate the relationship of DEGs and DAMs annotated with KEGG pathways, we first calculated the Pearson correlation coefficients of DEGs and DAMs. Pearson's correlation coefficient ≥ 0.8 was considered to be significantly correlated. Then, the significantly correlated DEGs and DAMs were mapped onto the KEGG pathway database to gather information about their shared pathways to better show their interaction. The KEGG pathway enrichment analysis showed that the DEGs and DRMs were commonly enriched in 18 KEGG pathways, such as the flavonoid, phenylpropanoid, linoleic acid metabolism, diterpenoid biosynthesis, riboflavin biosynthesis, and isoflavonoid biosynthesis ( Figure S10A). Among the 18 KEGG pathways, the isoflavonoid biosynthesis and phenylpropanoid biosynthesis pathways were significantly enriched. To reveal the correlation between the DEGs and DRMs involved in these two pathways, the networks between the DEGs and DRMs were constructed using the screening criteria of the absolute values of PCC ≥ 0.99 and p < 0.01 ( Figure S10B,C). The significance of flavonoid biosynthesis in RNAi-GiSRT2 hairy root lines was highlighted using transcriptome analysis, so this metabolic pathway was focused on in later analyses.
In the present study, the biosynthetic pathways of flavonoid and up-regulated DEGs (UDEGs) in RNAi-GiSRT2 hairy roots are summarized (Figure 4). In phenylpropanoid and flavonoid biosynthetic pathways, two PALs, fourteen CYPs, four OMTs, four GiCHSs, one GiCHI, and three GiFLSs were up-regulated. These results were well consistent with the enhanced accumulation of LCA and other flavonoids in RNAi-GiSRT2 lines. This suggested that the above genes are potential key ones regulating flavonoids biosynthesis in RNAi-GiSRT2 hairy roots.

GiLMT1 Is Involved in the Biosynthesis of LCA in G. inflata
To identify the genes involved in the biosynthesis of LCA, DEGs in the propo biosynthetic pathway ( Figure S4) were investigated. Up−regulated PALs, CYPs, CH CHI, F2H, and LMT may be involved in the biosynthesis of precursors ( Figure 4A). H we selected the LMT that was predicted to catalyze the next steps in the pathway, wh is more likely to be involved in LCA biosynthesis. We cloned the coding sequence Gi2.258 and named it GiLMT1. It encodes a protein which shares 83.60% identity w MsLMT ( Figure S11), the first cloned LMT from Medicago sativa L. [26]. The express level of GiLMT1 was higher in roots than in leaves, contrary to GiSRT2, suggesting opposite effects of GiLMT1 and GiSRT2 in LCA accumulation ( Figure S12).
To verify the function of GiLMT1, we carried out in vitro enzyme assays. The cru enzyme of GiLMT1 obtained by heterologous expression in E. coli was incubated w licodione and S−adenosyl methionine (SAM), the methyl donor. GiLMT1 could transfe methyl group to licodione and produce 2′−O−methyllicodione (P1) (Figure 5A,B). Mor ver, licodione is unstable and interconvertible with its isomer. GiLMT1 was also capa of catalyzing the licodione isomer, yielding the methylated product P2 (Figure 5A,B). T substrates and products were all confirmed using mass spectrometry ( Figure S13). Th results indicated that GiLMT1 could catalyze the methylation of licodione in vitro.

GiLMT1 Is Involved in the Biosynthesis of LCA in G. inflata
To identify the genes involved in the biosynthesis of LCA, DEGs in the proposed biosynthetic pathway ( Figure S4) were investigated. Up-regulated PALs, CYPs, CHSs, CHI, F2H, and LMT may be involved in the biosynthesis of precursors ( Figure 4A). Here we selected the LMT that was predicted to catalyze the next steps in the pathway, which is more likely to be involved in LCA biosynthesis. We cloned the coding sequence of Gi2.258 and named it GiLMT1. It encodes a protein which shares 83.60% identity with MsLMT ( Figure S11), the first cloned LMT from Medicago sativa L. [26]. The expression level of GiLMT1 was higher in roots than in leaves, contrary to GiSRT2, suggesting the opposite effects of GiLMT1 and GiSRT2 in LCA accumulation ( Figure S12).
To verify the function of GiLMT1, we carried out in vitro enzyme assays. The crude enzyme of GiLMT1 obtained by heterologous expression in E. coli was incubated with licodione and S-adenosyl methionine (SAM), the methyl donor. GiLMT1 could transfer a methyl group to licodione and produce 2 -O-methyllicodione (P1) (Figure 5A,B). Moreover, licodione is unstable and interconvertible with its isomer. GiLMT1 was also capable of catalyzing the licodione isomer, yielding the methylated product P2 (Figure 5A,B). The substrates and products were all confirmed using mass spectrometry ( Figure S13). These results indicated that GiLMT1 could catalyze the methylation of licodione in vitro.

Discussion
LCA is a characteristic flavonoid in G. inflata, a medicinal licorice species. De high economic value, the biosynthetic pathway and the molecular regulation mec are not clear. Identification of enzymes involved in LCA biosynthesis will help to struct the pathway in different organisms for the production of LCA. Furthermo covering factors that affect LCA accumulation would help to improve LCA co farmed licorice. To test the function of GiLMT1 in vivo, OE-GiLMT1 and RNAi-GiLMT1 transgenic hairy roots were generated and verified using qRT-PCR ( Figure 5C,D). Compared with hairy roots carrying pSuper-GFP empty vector, the OE-GiLMT1 lines displayed increased LCA contents (Figure 5E), which is similar to the case observed in RNAi-GiSRT2 lines. In contrast, the LCA contents in RNAi-GiLMT1 lines were significantly lower than those in the control lines ( Figure 5F). These results indicated that GiLMT1 is involved in the biosynthesis of LCA in G. inflata. However, more research is needed to determine whether GiLMT1 is a direct target of GiSRT2.

Discussion
LCA is a characteristic flavonoid in G. inflata, a medicinal licorice species. Despite its high economic value, the biosynthetic pathway and the molecular regulation mechanism are not clear. Identification of enzymes involved in LCA biosynthesis will help to reconstruct the pathway in different organisms for the production of LCA. Furthermore, discovering factors that affect LCA accumulation would help to improve LCA content in farmed licorice.

The Role of GiSRT2 in Balancing Growth and Resistance of G. inflata
A previous study found that when ginseng adventitious roots were treated with HDAC inhibitors SAHA or NaB, the MeJA-induced H3ac level was significantly increased, and the inhibition of HDAC activity improved MeJA-induced transcriptional activation of ginsenoside biosynthesis pathways (Lu and Hyun, 2021). Here we found that NIC treatments significantly increased the content of total flavonoids and LCA in G. inflata seedlings ( Figure 1C,D), which is consistent with the observation in the OE/RNAi-GiSRT2 transgenic hairy roots ( Figure 2D,E), suggesting that GiSRT2 is a negative regulator of flavonoid biosynthesis. Inhibition of GiSRT2 activity by NIC treatment or RNAi released the repression on flavonoid biosynthesis. Accumulation of flavonoids has been reported to be important protection against different environmental stresses such as UV, drought, etc. [27]. In white clover, different flavonoids induced in cells undergoing nodule organogenesis regulated local auxin levels either by promoting or repressing auxin breakdown and thus regulated root growth [28]. Since NIC treatment inhibited root growth but promoted flavonoid accumulation, our results indicated that GiSRT2 is an important regulator to balance plant growth and stress responses/accumulation of flavonoids. However, despite its inhibitory effect on plant growth, NIC is still a potential 'fertilizer' that could be used in the licorice field before root harvest, when certain biomass has been accumulated and a short NIC treatment may activate the LCA biosynthesis resulting in a better quality of licorice.

Determination of Gene Expression and Metabolic Changes in RNAi-GiSRT2 Hairy Roots through Transcriptomic and Metabolomic Techniques
Since the molecular mechanism of HDAC regulating flavonoid biosynthesis in plants is rarely reported, we carried out integrated transcriptomic and metabolomic analysis with RNAi-GiSRT2 transgenic hairy roots. Structural genes affecting specialized metabolism of G. inflata and metabolic pathways with significant enrichment were analyzed. A total of 4930 DEGs were identified in RNAi-GiSRT2. PAL, CYP73A, and CCOMT are associated with lignin biosynthesis. Twenty-three DEGs were annotated to the flavonoid biosynthesis pathway, including CHSs, CHI, F2H, FLS, and LMT. Consistently, the contents of 11 DAMs were mapped to flavonoid biosynthesis and accumulated more in RNAi-GiSRT2 hairy roots. Furthermore, the KEGG pathway enrichment analysis showed that the DEGs and DRMs were commonly enriched in 18 KEGG pathways involved in the phenylpropanoid, linoleic acid metabolism, diterpenoid biosynthesis, and riboflavin biosynthesis ( Figure S10A). These results not only lay the foundation for the excavation of functional genes related to the biosynthesis of LCA in G. inflata, but also provide an overall picture of sirtuins regulating plant specialized metabolism. Further study on sirtuins regulating the biosynthesis of different specialized metabolites could be carried out based on these omics data.

GiLMT1 Is Required for LCA Biosynthesis
In this work, GiLMT1 was found up-regulated in RNAi-GiSRT2 lines. The in vitro enzyme assays showed that GiLMT1 could add a methyl group to licodione, which is an intermediate step in LCA biosynthesis. Consistently, the transgenic hairy roots overexpressing GiLMT1 produced more LCA while the LCA contents were greatly reduced in RNAi lines. Both in vitro and in vivo data demonstrated the key role of GiLMT1 in the LCA biosynthesis process. Our work provides a key point that could be used in both molecular breeding of licorice with high LCA content, and in synthetic biology to produce LCA in different organisms. It is true that we cannot rule out the possibilities that other LMTs may be involved in LCA biosynthesis and more studies are needed to figure precisely out how GiSRT2 regulates the expression of GiLMT1.

Conclusions
In this study, we proved that an SIR class HDAC-specific inhibitor NIC could increase the content of flavonoids in G. inflata roots. We further cloned and characterized a NICtargeted HDAC gene, GiSRT2, which is predominantly located in the nucleus and negatively regulates the accumulation of flavonoids in hairy root transgenic lines. Combined analysis of metabolome and transcriptome with RNAi-GiSRT2 transgenic hairy roots revealed the potential mechanism underlying this process. The expression level of several structural genes of flavonoid biosynthesis pathways in RNAi-GiSRT2 hairy roots was up-regulated, among which GiLMT1 was proven to be required for the biosynthesis of LCA in G. inflata. This study provides new insights into the role of GiSRT2 in the regulation of the biosynthesis of specialized metabolites in G. inflata. It also provides evidence for the possibility of a novel metabolic engineering strategy to promote important specialized metabolite production in medicinal plants.

Supplementary Materials:
The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/cells12111501/s1, Figure S1: Analysis of flavonoid contents in non-transgenic hairy roots (WT), OE empty vector induced hairy roots (OE-CK), RNAi empty vector induced hairy roots (RNAi-CK); Figure S2: Expression pattern and subcellular localization of GiSRT2; Figure S3: Identification of OE-GiSRT2 and RNAi-GiSRT2 transgenic hairy roots; Figure S4: Proposed biosynthetic pathway of LCA in G. inflata; Figure S5: Gene expression analysis of RNA-seq samples; Figure S6: Gene functional enrichment analysis of DEGs; Figure S7: Analysis of differentially expressed transcription factors; Figure S8: Analysis of the metabolites detected by the metabolome; Figure S9: Overall analysis of the metabolomics data of RNAi-GiSRT2; Figure S10: Correlation analysis between DAMs and DEGs; Figure S11: Amino acid sequence alignment of GiLMT1 and MsLMT; Figure S12: Tissues expression pattern of GiLMT1 in G. inflata; Figure S13: Identification of licodione, its isomer and their methylated products by mass spectrum; Table S1: Primers used in this study.