Cyclocarya paliurus leaf samples from the same maturity (middle leaves) at different growth periods were collected from August 2020 to November 2020. The experiments were conducted on the middle leaves at each stage (8 M, harvested in August; 9 M, harvested in September; 10 M, harvested in October; and 11 M, harvested in November) (Figure 1A). The leaves were collected from the same three trees at different time periods with three biological replicates, and each tree was considered one biological repeat. Leaf samples were quickly frozen in liquid nitrogen and stored in a −80°C freezer for multiomics sequencing and analysis. Alpha-amylase (Cat# 8750) and α-amylase activity kits (Cat# BC0615) were purchased from Solarbio (Beijing, China). Neochlorogenic acid and vanillic acid assay kits were purchased from MedChemExpress (Shanghai, China).

Total polysaccharides, triterpenoids, and flavonoids content were determined by using UV-vis spectrophotometry according to the Chinese industry standards of SN/T4260-2015, NY/T3676-2020, and SN/T4592-2016, respectively (Supplementary Figure S1). The detection and quantitative analysis of hypoglycemic nutrient metabolites in the samples were performed by Metware (http://www.metware.cn/) based on comparison with a self-built database of purified metabolite standards and public database. Briefly, the freeze-dried leaves were crushed by a grinding mill (MM 400, Retsch) at 30 Hz for 1.5 min. One hundred milligrams of powder was dissolved in 1.2 mL of 70% methanol, vortexed for 30 s, and placed at 4°C overnight. The supernatant was obtained by centrifugation (12,000 rpm, 10 min), filtered through a 0.22-μm membrane, and then analyzed using a UPLC/MS/MS system (19). The operating conditions of UPLC were as follows: column, Agilent SB-C18 column (1.8 μm, 2.1 × 100 mm); mobile solvent A (0.1% formic acid in water) and mobile solvent B (0.1% formic acid in acetonitrile); flow rate, 0.35 mL/min; injection volume, 4 μL. The operating parameters of the MS/MS were as follows: ion spray voltage, 5,500 V (+)/4,500 V (–); ion source gas, gas I (50 psi), gas II (60 psi), and curtain gas (25 psi); and ion source temperature, 550°C. All metabolites were identified qualitatively by MetWare database (MWDB), and isotope signals, repeated signals containing K⁺ ions, Na⁺ ions and NH4+ ions, as well as repeated signals of fragments of other substances with larger molecular weights were removed during analysis based on the secondary spectrum information. Metabolites were quantified by multiple reaction monitoring (MRM) using triple quadrupole mass spectrometry. After obtaining the metabolite spectrum analysis data of different samples, the peak area integration was performed on the mass spectrum peaks of all substances, and the integral correction was performed on the mass spectrum peaks of the same metabolite in different samples (20). Mass spectrum data was conducted with Analyst software (version 1.6.1) (21). The characteristic ions of each substance were screened by triple quadrupole, and the signal intensity of the characteristic ions was obtained in the detector. The mass spectrometry data was analyzed using MultiaQuant software, and the chromatographic peaks were integrated and corrected. The peak area of each chromatographic peak represents the relative content of the corresponding substance. Principal component analysis (PCA) was carried out by statistics function prcomp within R (www.r-project.org) with the parameter scale = True, and hierarchical cluster analysis (HCA) was performed by R package pheatmap. Orthogonal partial least squares-discriminant analysis (OPLS-DA) model was established by using function OPLSR.Anal within R package MetaboAnalystR (22, 23). Significantly regulated metabolites between groups were determined (log2-fold change ≥ 2 and variable importance in project ≥ 1) and then enriched using the KEGG database.

Alpha-amylase (EC 3.2.1.1) and α-glucosidase (EC 3.2.1.20) belong to glycosidases (EC 3.2.1) are involved in hydrolysis of starch into sugars and disaccharides, and inhibition of these enzymes activity can effectively lower the blood glucose level (24). Acarbose, as a weak and low-selective inhibitor, is the dominant drug in the field of diabetes by slowing down the absorption of glucose through the inhibition of α-amylase, howbeit with serious side effects (25). Inhibition of α-amylase activity provide an essential way for the treatment of type 2 diabetes mellitus (T2DM), which can be used as the basis for the development of new hypoglycemic drugs in the future (26). To detect the α-amylase inhibition activity, a mixture of 37.5 μL of α-amylase solution (0.1 mg/mL) and 5.0 μL of α-amylase inhibitor (neochlorogenic acid or vanillic acid) was first incubated at 40°C for 5 min, and then 75 μL of solution I (Cat#BC0615, Solarbio) was added. The reaction was stopped by heating for 10 min at 100°C after the addition of 37.5 μL of solution II (Cat#BC0615, Solarbio). The reaction solution was cooled to room temperature and then measured at 540 nm using a microplate reader (OPTIMA S/N413-3915). Acarbose and boiled α-amylase solutions were used as the positive and blank controls, respectively. A control reaction was used, in which the α-amylase solution was replaced with buffer solution. A negative reaction was used, in which the inhibitor was replaced with the buffer solution. Inhibitory concentrations of 50% (IC₅₀) were analyzed by GraphPad Prism using the specific calculation formula (Y = Bottom + (Top-Bottom)/(1 + 10^∧((LogIC₅₀-X)^*HillSlope))). The inhibition percentage of α-amylase was calculated by the following formula:

where A₁ = activity with negative control (without inhibitor), A₂ = activity with blank control (with boiled α-amylase solution), A₃ = activity with inhibitor (acarbose, neochlorogenic acid, or vanillic acid), and A₄ = activity with control reaction (without α-amylase solution). One unit of enzyme inhibition was expressed by the weight of the IC₅₀ value per milliliter.

The preparation of total RNA was completed with the RNeasy plant mini kit (Tiangen Bio, Beijing, China). Construction of libraries for the rRNA-depleted stranded RNA-seq and small RNA was carried out with the TruSeq Stranded Total RNA Prep kit (Illumina, San Diego, USA) and the TruSeq® Small RNA sample preparation guide (Illumina, San Diego, CA), respectively. HISAT2 was used for aligning trimmed sequences to the reference genome (27). StringTie (version 1.3.1) was used to calculate the FPKM values of lncRNAs and mRNAs (28), miRNAs were detected using mirDeep2 (version 2.0.5), and miRNA expression was calculated from miRBase. TargetFinder (version 1.6) was used to predict miRNA target genes (29). CIRI (version 1.1.1) was used to identify and annotate the de novo circRNAs (30). Differential expression analysis of RNAs was performed (fold change ≥2.0, p-value < 0.05, and FDR <0.05) and then enriched by KEGG (31).

Coexpression analysis was performed by a threshold of PCC ≥ 0.9 and a p-value < 0.01 (32–35). The ceRNA relationship pairs were obtained through the targeting relationship of miRNAs based on the number of miRNAs > 5, FDR ≤ 0.01, and a p-value < 0.01. Based on the ceRNA results, the relationship pairs of all differences of lncRNA-miRNA-mRNA and circRNA-mRNA-mRNA were obtained and visualized using Cytoscape (version 3.7.2) (36).

The transcriptome data were normalized using the normalize quantiles function in the R software package, and all differentially expressed genes (DEGs) between samples were used for further WGNCA analysis (35). The eigengenes of each module were calculated, and then cluster analysis was performed on the modules. WGCNA was performed as follows: minimum module size, 30; soft power, 13; and merge cut height, 0.15 (Supplementary Figure S2). Eigengene values of modules related to polysaccharides, triterpenoids, and flavonoids were calculated. The hub genes within significant modules were considered with a coefficient > 0.6 and a p-value < 0.05. All genes and module eigengene genes in the module were analyzed by expression level, GO, and KEGG analyses. Cytoscape (version 3.7.2) was used to visualize the coexpression network of ME genes in the module of interest (36).

Total RNA was completed by using the TransZol Up Plus RNA Kit (Transgen Biotech, Beijing). First-strand cDNA was synthesized by TransScript One-Step gDNA Removal and cDNA Synthesis SuperMix (Transgen Biotech, Beijing) followed by analysis using the SYBR PrimeScript RT–PCR kit (Takara, Beijing). The specific primers for qRT–PCR were designed using Primer 5.0 (Supplementary Table S1). The relative level of mRNA expression was calculated according to the 2^−ΔΔCt method and GAPDH was used as the reference gene (37).

To identify the major metabolites in C. paliurus leaves at different harvest stages, we determined the total polysaccharides, triterpenoids, and flavonoids content (Figures 1B–D; Supplementary Tables S2–S4). The results showed that the total flavonoids content increased significantly during the development of C. paliurus leaves: from 4.09 mg/g in August (8 M) to 9.34 mg/g in October (10 M) (Figure 1B). The total triterpenoids content of 10 M (51.89 mg/g) was 1.90-, 1.80-, and 2.19-fold higher than that of the 8 M (27.33 mg/g), 9 M (28.79 mg/g), and 11 M (23.66 mg/g) groups, respectively (Figure 1C). In addition, the total polysaccharides content in C. paliurus leaves reached its maximum in the 10 M group but started to decrease significantly in the 11 M group (Figure 1D). Therefore, the findings confirmed that October may be conducive to the accumulation of hypoglycemic nutrient metabolites. To elucidate the potential chemical basis for the nutritional quality at different harvest stages, widely targeted metabolomic analysis was conducted using the UPLC–MS/MS detection platform. PCA and HCA analysis indicated that the metabolic profiles in C. paliurus leaves at different harvest stages exhibited a significant difference (Figures 2A,B). Based on the results from the volcano plot, among the 751 metabolites detected there were: 363 differentially accumulated metabolites (DAMs) (203 up- and 160 downregulated) between the 8 M and 9 M, 286 DAMs (187 up- and 99 downregulated) between the 8 M and 10 M, 216 DAMs (106 up- and 110 downregulated) between the 8 and 11 M, 295 DAMs (182 up- and 113 downregulated) between the 9 and 10 M, 373 DAMs (170 up- and 203 downregulated) between the 9 and 11 M, and 323 DAMs (124 up- and 199 downregulated) between the 10 and 11 M (fold change ≥ 2 and VIP ≥ 1) (Supplementary Figure S3; Supplementary Table S5). The Venn diagram analysis revealed that 12 DAMs were shared among all samples, whereas 16, 11, 9, 14, 10, and 8 different DAMs were unique to 8M_vs_9 M, 8M_vs_10 M, 8M_vs_11 M, 9M_vs_10 M, 9M_vs_11 M, and 10M_vs_11 M, respectively, indicating that the quality changes of C. paliurus leaves may be related to the regulation of these unique DAMs in multiple pathways (Figure 2C). Additionally, these DAMs were classified into nine groups, with the largest number of metabolites falling under flavonoids (124, 20.74%), followed by phenolic acids (100, 16.72%), lipids (77, 12.88%), amino acids, and derivatives (61, 10.20%), organic acids (59, 9.87%), and terpenoids (54, 9.03%) (Figure 2D). The contents of phenolic acids (vanillic acid, 1.17-fold increase; 4-hydroxybenzoic acid, 1.29-fold increase; anthranilic acid, 1.59-fold increase; caffeic acid, 2.23-fold increase) and flavonoids (dihydromyricetin, 12.12-fold increase; kaempferol-3-O-neohesperidoside, 1.37-fold increase; quercetin-3-O-sambubioside, 2.11-fold increase) were significantly higher in the 9 M group than in the 8 M group (Figure 3A; Supplementary Table S6). In the comparison of 8 vs. 10 M, most secondary metabolites (2,3-dihydroxy-3-methylbutanoic acid, 17.60-fold increase; dihydromyricetin, 12.43-fold increase; cyclocaric acid B, 3.22-fold increase; cyclocaric acid A, 2.28-fold increase; neochlorogenic acid, 1.06) were upregulated in 10 M compared to 8 M (Figure 3B; Supplementary Table S6). In the comparison of 10 M vs. 11 M, the metabolite profiles associated with flavonoids (tricetin, 15.56-fold decrease; kaempferol-3-O-rutinoside, 14.31-fold decrease), phenolic acids (2-methoxybenzoic acid, 10.98-fold decrease; acetovanillone, 10.18-fold decrease; neochlorogenic acid, 1.33-fold decrease), and nucleotides and derivatives (inosine, 12.09-fold decrease; 2-deoxyribose-1-phosphate, 15.65-fold decrease) were significantly downregulated in the 11 M group compared with the 10 M group (Figure 3C; Supplementary Table S6). The number of DAMs between the 9 M and 11 M groups was greatest among these comparison groups, and the number of downregulated DAMs (203) was 1.19-fold higher than that of upregulated DAMs (170), indicating that the quality of C. paliurus leaves may decline from the 9 M to the 11 M stage (Figure 3D; Supplementary Table S6). Considering the accumulation of metabolites, a suitable harvesting strategy for C. paliurus leaves could be in September or October each year.

To confirm the metabolomic results and also to develop potential alpha-amylase inhibitors from C. paliurus leaves, an α-amylase inhibitory activity assay was conducted on seven differentially accumulated phenolic acid metabolites selected based on the metabolic changes and differential analysis, including vanillic acid, 4-hydroxybenzoic acid, anthranilic acid, caffeic acid, 2-methoxybenzoic acid, acetovanillone, and neochlorogenic acid. The α-amylase inhibitory activity results showed that neochlorogenic acid, vanillic acid, and acarbose reduced the stability of the α-1,4-glucoside bond with IC₅₀ values of 0.23, 0.33, and 0.03 mM, respectively. We also found that the highest inhibition of neochlorogenic acid and vanillic acid was 61.74 and 57.03% at a concentration of 1.0 mM, which was lower than that of acarbose at 1.0 mM (85.67%) (Figure 4). The results showed that neochlorogenic acid and vanillic acid have a relatively good inhibitory effect on α-amylase, indicating that they may play a role in hypoglycemic functions.

To identify the critical genes associated with hypoglycemic nutrient metabolites in C. paliurus leaves, the whole transcriptome was further analyzed. A total of 8,805 lncRNAs, 3,2164 mRNAs, 899 circRNAs, and 496 miRNAs (43 known miRNAs, 453 novel miRNAs, and 4,125 target genes) were identified in the 8–11 M groups, respectively (Figure 5A; Supplementary Figure S4A). Of those, 21,411 (84.28%) of 25,405 new genes were successfully annotated to nine public databases, including COG (6,654, 26.19%), KEGG (14,938, 58.79%), GO (17,461, 68.73%), KOG (11,800, 46.45%), Pfam (16,626, 65.44%), SwissProt (15,067, 59.31%), TrEMBL (21,343, 84.01%), eggNOG (18,139, 71.40%), and NR (21,313, 83.89%) (Supplementary Figure S4B; Supplementary Table S7). Based on the expression profiles and differential analysis, a total of 15,951 differentially expressed (DE) mRNAs were identified, among which 7,110 and 8,841 DE-mRNAs showed upregulation and downregulation, respectively (Supplementary Figure S4C). Among these DE-mRNAs, the largest number of DEGs fell under 8M_vs_11 M (2,828 upregulated and 4,064 downregulated), followed by 9M_vs_11 M (2,623 upregulated and 3,253 downregulated), and 8M_vs_9 M (1,226 upregulated and 1,184 downregulated), and the number of downregulated genes (4,064) was 1.44-fold higher than that of upregulated genes (4,064) from the 8M_vs_11 M group (Figure 5B; Supplementary Figure S5). Enrichment analysis showed that these DE-mRNAs were mainly associated with GO:0006952~defense response, GO:0045490~pectin catabolic process, phenylpropanoid biosynthesis, plant hormone signal transduction, glucuronate interconversions, fructose metabolism, diterpenoid biosynthesis, and fatty acid degradation (Figure 5C). In the comparison of 8 vs. 11 M, most downregulated DEGs were enriched in the starch and sucrose pathway (CpsacA, newGene_11051; CpSUS, newGene_11769; CpGH12, newGene_18867; CpHK, newGene_16922), amino sugar and nucleotide sugar metabolism (CpgalE, newGene_7290; CpglgC, newGene_2262; CpUSP, newGene_15611; CpPMM, newGene_21902), and fructose and mannose metabolism (CpPFP, newGene_9652; CpGMPP, newGene_5583; CpMAN, newGene_11578) (Supplementary Figure S6A). Notable decreases in transcriptome expression from the 9 M to 11 M group included terpenoid biosynthesis (CpCYP76F14, newGene_34701; CpLUS, newGene_10624; CpNES1, newGene_8992; CpKAO, newGene_27209) and flavonoid biosynthesis (CpCYP3A, newGene_20322; CpANR, newGene_20574; CpHCT, newGene_22094; CpDFR, newGene_25488; CpANS, newGene_35209) (Supplementary Figure S6B). We hypothesize that most genes involved in secondary metabolites are significantly downregulated in November collected C. paliurus leaves compared with those collected in August and September. For miRNAs, 3,703 (89.77%) of 4,125 target genes were successfully annotated to nine public databases, including COG (1,460, 35.39%), KEGG (2,760, 66.91%), GO (3,177, 77.02%), KOG (2,112, 51.20%), Pfam (3,051, 73.96%), SwissProt (2,828, 68.56%), eggNOG (3,275, 79.39%), and NR (3,698, 89.65%) (Supplementary Figure S7; Supplementary Table S8). Of these, 722 DE-miRNAs were identified by pairwise comparison, of which 404 were upregulated and 318 were downregulated (Supplementary Table S9). Enrichment analysis of DE-miRNA target genes showed that they were mainly enriched in plant–pathogen interactions, plant hormone signal transduction, glycosphingolipid biosynthesis, terpenoid biosynthesis, starch and sucrose metabolism, and phenylpropanoid biosynthesis (Supplementary Figure S8). Additionally, we also obtained 12 DE-circRNAs, including 9 upregulated and 3 downregulated circRNAs. For example, DE-circRNA Chr10:10:9049396|9051972 was upregulated with CpNAC62 (newGene_16052), which is involved in the defense response; DE-circRNA Chr03:22680395|22837830 was positively correlated with CpGME (newGene_37112), which is related to amino sugar and nucleotide sugar metabolism (Supplementary Figure S9).

A coexpression network was constructed, resulting in 1,048,575 lncRNA-mRNA, 296,770 circRNA-mRNA, 189,105 circRNA-lncRNA, and 589 miRNA-circRNA networks. A ceRNA network was generated with 76 nodes based on the coexpression relationship, including 16 lncRNAs, 32 circRNAs, and 28 mRNAs. Enrichment analysis implied that the relationship pairs from DE-ceRNAs mainly participated in alpha-linolenic acid metabolism, MAPK signaling pathway, phenylpropanoid biosynthesis, and diterpenoid biosynthesis (Supplementary Figure S10). The circRNA-miRNA-mRNA network indicated that gma-miR-156k, one of the key miRNAs, inhibited the expression of CpPIP5K (newGene_1012) by regulating Chr02:8978909|9035691. Moreover, circRNAs Chr08:5445135|5475645 and CpMFP2 (newGene_21820) were positively correlated with gma-miR-156k, which is associated with alpha-linolenic acid metabolism (Figure 5D). The lncRNA-miRNA-mRNA network indicated that rco-miR156b established common connections with six lncRNAs and five mRNAs. LncRNA MSTRG.26466.1 acted as a sponge of rco-miR156b to regulate the target gene CpGBE1 (newGene_7975), which is related to starch and sucrose metabolism (Figure 5E).

To identify the hub genes associated with polysaccharides, triterpenoids, and flavonoids, we performed WGCNA for the RNA-seq data of 12 samples. We identified 25 modules (comprising of 24 coexpression modules and one module with uncorrelated genes), and the number of genes ranged from 35 to 2,938 (Figure 6A). Notably, the green module was found to have a significantly high association with polysaccharides (r = 0.61 and p-value = 0.03) (Figure 6B). Enrichment analysis showed that genes in the green module were significantly involved in galactose metabolism, sucrose metabolism, and fructose metabolism, including CpPMM (newGene_21902), CpMan (newGene_25062), CpFK (newGene_21700), CpSUS (newGene_17938), CpTPS (newGene_30013), CpbglX (newGene_17774), CpGN1 (newGene_6921), CpPYG (newGene_13377), and CpglgC (newGene_2262) (Figure 6C; Supplementary Figure S11). Furthermore, we found that two modules, dark-red (r = −0.83 and p-value = 8E-4) and green-yellow (r = −0.82 and p-value = 0.001), were significantly and negatively correlated with polysaccharides (Figure 6B). CpsucA (newGene_35015), CpsucB (qql_045894), CpENO (qql_090117), and CpCS (qql_024905) in the green-yellow module were annotated to be associated with the citrate cycle and glycolysis (Supplementary Figure S12). We hypothesis that these genes can reduce the production of glucose and energy by playing an important negative regulatory role in polysaccharide metabolism. Another light-yellow module was detected, which tended to be involved in flavonoids (r = 0.64 and p-value = 0.03) (Figure 6B; Supplementary Figure S13). We captured two hub genes (CpHCT, newGene_24442; Cp4CL, newGene_25920), which shared 29 and 44 edges with other candidate genes. In addition, we also identified 17 TFs in the black module and found several TFs that are highly positively correlated with triterpenoids, including two MYB-related genes (newGene_2469, newGene_19602) and one WRKY gene (CpWRKY1, newGene_8245) (Figure 6D; Supplementary Figure S14). To further confirm the reliability of the RNA-Seq analysis, qRT–PCR was performed on five DEGs, including CpFK, CpSUS, CpHCT, Cp4CL, and CpWRKY1. The results of qRT–PCR were similar to the FPKM values from RNA-Seq, demonstrating the high reliability of the RNA-seq data (Supplementary Figure S15).