Fresh kiwifruit samples from five varieties were selected: GC (Actinidia chinensis ‘Guichang’, Zhaotong), WL (Actinidia chinensis, Wild Kiwifruit, Zhaotong), LC (Actinidia arguta ‘Maolvfeng’, Dandong), DY (Actinidia arguta, Danyang Flat Kiwifruit, Dandong), and SG (Actinidia chinensis ‘SunGold’, New Zealand). Fruits from all varieties were harvested at a commercially ripe stage. For this comparative study, all post-harvest handling and sampling were performed under identical conditions. The fruits were gently washed with clean water to remove surface soil and impurities, then air-dried on sterile gauze or absorbent paper.

All grinding tools, including mortar and pestle, were wrapped in kraft paper and sterilized using an autoclave at 121°C for 20–30 minutes. Using a sterile scalpel, the skin of each fruit was carefully removed, and only the flesh was used for subsequent analysis. For each variety, flesh from multiple fruits was combined and ground into a homogeneous mixture to create one biological replicate. Three independent biological replicates were prepared per variety. A portion of the ground pulp was tightly wrapped in aluminum foil to form spherical samples, which were then placed into pre-cooled centrifuge tubes and appropriately labeled.

The samples were rapidly frozen by immersing the tubes in liquid nitrogen for 2–5 minutes to ensure complete solidification. Finally, all samples were stored together in a –80°C ultra-low temperature freezer for further metabolic analysis.

The reducing sugar content in kiwifruit samples was quantified using the 3,5-dinitrosalicylic acid (DNS) method (Lam et al., 2021). Approximately 0.1 g of fruit tissue from each variety was homogenized to 0.8 mL of 80% ethanol using an ice bath, and the mixture was transferred to a centrifuge tube. The final volume was adjusted to 1.5 mL with 80% ethanol.

The tubes were sealed and incubated in a water bath at 50°C for 20 minutes, mixing every 2 minutes. After cooling, the volume was readjusted if necessary, and the extracts were centrifuged at 12,000 rpm for 10 minutes. The supernatant was collected for analysis. A 100 µL aliquot of the supernatant was mixed with 100 µL of DNS reagent in a microtube, heated at 95°C for 10 minutes, and then cooled rapidly. After adding 1 mL of distilled water, 200 µL of the mixture was transferred to a 96-well plate, and the absorbance was measured at 500 nm using a microplate reader. A standard curve was generated using glucose standards, and the reducing sugar content was calculated according to the formula:

where ΔA is the corrected absorbance, W is the sample weight (g), and D is the dilution factor.

The soluble sugar content was determined using the anthrone-sulfuric acid method (Liu et al., 2025). Briefly, approximately 0.1 g of fruit tissue from each variety was homogenized in 80% ethanol, with the final extract volume adjusted to 1.5 mL. The extracts were incubated at 50°C for 20 minutes, centrifuged, and the supernatant was collected. For the assay, 25 µL of the supernatant was mixed with 75 µL of distilled water and 250 µL of anthrone reagent. Then, 250 µL of concentrated sulfuric acid was added slowly to the mixture. The tubes were heated at 95–100°C for 10 minutes, cooled, and the absorbance was measured at 620 nm using a microplate reader. A standard curve was generated using glucose, and the soluble sugar content was calculated according to the following formula:

where ΔA is the corrected absorbance, W is the sample fresh weight (g), and D is the dilution factor.

The DPPH (1,1-diphenyl-2-picrylhydrazyl) radical scavenging activity was measured to evaluate the antioxidant capacity of the kiwifruit samples (Wojtunik et al., 2014). Approximately 0.1 g of kiwifruit sample was ground, passed through a 40-mesh sieve, and extracted with 1 mL of 80% methanol using ultrasonic assistance at 60°C for 30 minutes. After centrifugation at 12,000 rpm for 10 minutes, the supernatant was collected. For the assay, 150 µL of the extract was mixed with 150 µL of DPPH working solution. The mixture was incubated in the dark at 25°C for 30 minutes, centrifuged, and the absorbance of the supernatant was measured at 517 nm. A standard curve was prepared using Trolox, and the radical scavenging activity was calculated as follows:

where W is the sample fresh weight (g), and D is the dilution factor.

A widely targeted metabolomic approach was employed to profile chemical diversity of terpenoids across the five kiwifruit varieties using an Ultra Performance Liquid Chromatography-Tandem Mass Spectrometry (UPLC-MS/MS) platform (Wuhan Metware Biotechnology Co., Ltd.). Three independent biological replicates per variety were analyzed. The analysis was performed on a system consisting of an ExionLC™ AD UPLC (SCIEX) for chromatographic separation coupled to a triple quadrupole-linear ion trap (QTRAP^® 6500+) mass spectrometer (SCIEX) for detection. Fresh fruit samples were freeze-dried, ground to a powder, and 30 mg of the powder was extracted with 1500 μL of a pre-cooled 70% methanol solution containing internal standards (CAS:14091-11-3; 2-chlorophenylalanine, 1PPM mg/L) for quality control and relative quantification. The extracts were vortexed, centrifuged, and the supernatant was filtered through a 0.22 μm membrane for analysis. A pooled quality control (QC) sample, prepared by combining equal aliquots from all individual extracts, was injected at regular intervals, to monitor instrument stability and correct for signal drift. Chromatographic separation was performed on an Agilent SB-C18 column (2.1 mm × 100 mm, 1.8 μm) using a gradient of water (A) and acetonitrile (B), both containing 0.1% formic acid, at a flow rate of 0.35 mL/min and a column temperature of 40°C.

Mass spectrometric detection was carried out on a triple quadrupole (QQQ) instrument equipped with an electrospray ionization (ESI) source operating in both positive and negative modes. Data acquisition was performed in Multiple Reaction Monitoring (MRM) mode, on a triple quadrupole mass spectrometer, with method parameters established from the Metware Database (MWDB). Data were acquired in both positive and negative electrospray ionization modes (ESI+/-) with ion spray voltages of 5500 V and -4500 V, respectively, a source temperature of 500°C, and gas parameters set at 50 psi (GSI), 60 psi (GSII), and 25 psi (curtain gas). For each metabolite, unique MRM transitions were established by optimizing the declustering potential and collision energy to select a specific precursor ion in the first quadrupole (Q1) and its most abundant and characteristic product ion in the third quadrupole (Q3), with a medium setting for collision-induced dissociation nitrogen gas. A scheduled MRM algorithm was employed, monitoring specific metabolite transitions only during their predefined retention windows to ensure sufficient data point acquisition across chromatographic peaks for precise relative quantification based on integrated peak areas, normalized to the internal standards and sample weight (Chen et al., 2013; Rao et al., 2022). The molecular formula, precursor (Q1) and product (Q3) ions for MRM transitions, molecular characteristics, ionization mode, confidence level of identification, and CAS registry number of all detected terpenoids are represented (Supplementary Table S1). Metabolite identification and qualification data were achieved by matching the acquired MS/MS spectra and retention times against the proprietary Metware Database (MWDB).

Multivariate statistical analyses were conducted to explore the metabolic variation among the five kiwifruit varieties. The relative quantification data from the metabolomic profiling were initially standardized using unit variance scaling. Principal component analysis (PCA) was performed to visualize the clustering of the kiwifruit samples. Variable Importance in Projection (VIP) scores from the OPLS-DA model were calculated, and metabolites with a VIP score > 1.0 were considered significant contributors to group discrimination (Thévenot et al., 2015).

Differential metabolites between comparison groups were identified by applying a dual-threshold criterion: a fold change of |log2FC| ≥ 1.0 and a statistically significant p-value (< 0.05) from a student’s t-test, adjusted for multiple comparisons using the false discovery rate (FDR) (Colquhoun, 2014). Hierarchical cluster analysis (HCA) was performed on the normalized data to visualize the accumulation patterns of metabolites across all kiwifruit samples. All statistical analyses and visualizations, including PCA and HCA, were performed using R software (version 4.1.2) (Chanana et al., 2020). For the analysis of biochemical traits (e.g., antioxidant capacity, sugar content), Fisher’s Least Significant Difference (LSD) test was conducted using Statistix 8.1, with significance defined at p< 0.05. EVenn web-based online (https://www.bic.ac.cn/EVenn/#/) free platform was employed to perform Venn analysis (Yang et al., 2024).

Liquid chromatography tandem mass spectrometry (LC-MS/MS) analysis revealed a diversity of terpenoids across the five kiwifruit varieties studied. In total, 309 distinct terpenoid compounds were identified and divided into major terpenoid classes such as 82 triterpenes, 79 sesquiterpenoids, 75 monoterpenoids, 57 diterpenoids, 10 terpene, and 6 triterpene saponins, underscoring the complex phytochemical diversity of kiwifruit (Supplementary Table S1).

The visual analysis of the five kiwifruit varieties reveals significant morphological diversity in terms of fruit outer skin color, size, and shape (Figure 1A). The GC and WL fruits are oval in shape. The LC and DY are intense green outer skin, however, the LC fruit is oval while the DY fruit is flat. The SG is brown in color. The soluble sugar content results revealed that the SG fruits exhibited higher reducing sugar (13.64 mg/g), non-reducing sugar (300.4 mg/g), and total soluble sugar content (314.04 mg/g), which was approximately 2.4 to 6 times greater than the other varieties (Figures 1B–D). The GC variety showed the second-highest level (131.72 mg/g), while the remaining three varieties (LC, DY, and WL) showed considerably lower and relatively similar total sugar contents, ranging from 52.76 to 57.47 mg/g. Similar pattern was observed in non-reducing sugars, which constituted the dominant fraction of soluble sugars in all varieties, with SG and GC again displaying the highest concentrations. Reducing sugar content was less variable but was also highest in the SG variety. This stark contrast highlights the likely impact of genetic background on sugar accumulation, with the commercial SG variety indicating a superior sweetness potential.

The quantitative analysis, based on the sum of all chromatographic peak areas, revealed significant varietal differences in the total terpenoid content (Figure 1E). The SG variety from New Zealand and the DY flat kiwifruit exhibited the highest overall terpenoid abundance (Figure 1E). In contrast, the WL and the GC variety, both from Zhaotong, displayed substantially lower total terpenoid contents (Figure 1E). These findings suggest that genetic background (variety) and geographical origin along with possible environmental and agronomic factors are key factors influencing the terpenoid biosynthetic pathway in kiwifruit.

Principal Component Analysis (PCA) of the terpenoid profiles revealed clear separations among the five kiwifruit varieties, indicating that their overall terpenoid compositions are distinct and characteristic (Figure 2A). The first two principal components (PC1 and PC2) explained a cumulative 75.65% of the total variance (54.81% and 20.84%, respectively). The clear separation indicates that each variety possesses a unique and characteristic terpenoid composition. Compound-wise PCA showed significant variation among terpenoids, the PC1 accounted for 64.53% while the PC2 accounted for 21.81% of variation (Figure 2B). Most of the terpenoids were clustered near the intersection point of x-axis and y-axis, while a few numbers of compounds scattered across the PCA-plot, showing dissimilar characteristics among terpenoids (Figure 2B). This structured grouping demonstrates that the terpenoid metabolome is a strong chemotaxonomic marker capable of distinguishing kiwifruit genotypes. Correlation analysis showed a strong positive correlation in terpenoid composition among biological replicates of the same variety (Figure 2C). Most intra-variety correlation coefficients exceeded 0.90, indicating high reproducibility and a homogeneous terpenoid profile for each genotype. In contrast, inter-variety comparisons generally showed low or negative correlations, indicating that each kiwifruit variety possesses a distinct terpenoid fingerprint.

Hierarchical Clustering Analysis (HCA) was performed on the 198 terpenoids that were identified as significantly altered (VIP > 1.1). The heatmap (Figure 3) revealed clear varietal-specific differences in terpenoid accumulation. The HCA showed that DY variety was high in abundance of ursane-type triterpenes, such as 2,3-Dihydroxy-12-ursen-28-oic acid, pomolic acid, and jujubogenin present than in other varieties (Figure 3). The WL and LC varieties formed a cluster largely due to their shared, high levels of oleanane-type triterpenes, such as alphitolic acid, gypensapogenin F, and crataegolic acid, which were consistently elevated in WL and LC compared to GC and SG. In contrast, the SG variety was separated by a unique combination of moderate to high levels of specific triterpenoids like ursolic acid, oleanolic acid, and asiatic acid. The cultivated GC consistently shows low concentrations of triterpenoids that were elevated in the other varieties. This general suppression of triterpenoid biosynthesis was a key factor in its clear separation from its wild (WL) and the other varieties. This HCA analysis determines that the terpenoid landscape is highly dependent on the kiwifruit genotype and origin.

The Venn diagram analysis revealed distinct profiles of terpenoids among different comparison groups of five kiwifruit varieties (DY, SG, LC, GC, WL) (Figures 4A, B). A total of 38 compounds were commonly present in all five comparison groups, the DY vs GC showed highest 235 number of identified compounds while the GC vs SG showed lowest 151 number of terpenoids (Figure 4A). Similarly, in second Venn diagram, a total of 23 compounds were commonly present in all five comparison groups, the LC vs SG showed highest 203 number of identified compounds followed by SG vs WL 199 while the LC vs DY showed lowest 162 number of terpenoids (Figure 4B). The comparison group LC vs DY revealed maximum number of 23 unique compounds followed by LC vs GC 11 (Figure 4B). This indicates that a set of compounds is common among different comparison groups, while each group possesses a unique terpenoid fingerprint, suggesting significant terpenoids diversity among kiwifruit varieties.

Pairwise comparison of terpenoid profiles were performed to evaluate the significantly altered (Variable Importance in Projection, VIP > 1.0 and fold-change, FC > 1) terpenoids among varieties (Figure 4C). The comparison between DY and WL showed the most pronounced variation, with 208 terpenoids significantly upregulated in DY and only 16 upregulated in WL. Similarly, comparisons involving the high-terpenoid varieties (DY and SG) against the low-terpenoid ones (GC and WL) consistently showed high number of upregulated metabolites. In contrast, the comparison between the two low-terpenoid varieties, GC and WL, showed a more balanced alteration (94 up in GC, 80 up in WL). Notably, the comparison between GC and SG revealed that 124 terpenoids were significantly higher in SG, while only 27 were higher in GC, highlighting the distinct phytochemical signature of the commercial ‘SunGold’ variety (Figure 4C). These results confirm that beyond the total content, the specific composition and abundance of individual terpenoids are highly variety specific.

The percentage abundance of 22 major terpenoids showed differences among kiwifruit varieties (Figure 5; Supplementary Table S4). The SG variety showed higher percentage abundance of specific triterpenic acids, such as cannabifolin C, asiatic acid, and madasiatic acid, which were about 1.5 to 2.7 times higher than other varieties. This suggests a distinct biosynthetic pathway favoring these compounds in the SG genotype. In contrast, the LC and DY varieties are rich in pentacyclic triterpenes such as gypensapogenin F, 2-hydroxyursolic acid, and alphitolic acid, though DY exhibited higher concentrations across these compounds. Notably, pomolic acid and its derivative, 2,3-dihydroxy-12-ursen-28-oic acid, were exceptionally abundant in the DY variety, serving as a key chemical signature (Figure 5).

The WL displayed diverse terpenoid profile, with high levels of alphitolic acid, gypensapogenin F, and javanicolide C, indicating a complex and potent terpenoid backbone. Its cultivated counterpart from the same region, GC, was starkly different, showing a general suppression of most triterpenic acids but a unique, high accumulation of 8-(Beta-D-Glucopyranosyloxy)-2,7-Dimethyldec-2-enedioic acid, a glycosylated terpenoid derivative. This obvious difference among the wild and cultivated types from the same geographical origin highlights the significant impact of possible breeding and selection effect on secondary metabolite production. Overall, the results showed that each kiwifruit variety possesses a unique terpenoid “fingerprint,” with implications for their sensory properties, nutritional value, and potential bioactivity.

The quantitative analysis of terpenoid compounds revealed significant variations in their abundance across the sample groups. Several terpenoids demonstrated significant alteration, as indicated by their log2 Fold Change (log₂FC) values, which were supported by statistically significant p-values (Figure 6). The varieties SG and DY showed significantly upregulation (Log₂FC > 8) of most of the terpenoids compounds as compared to other three varieties. These results highlight that the terpenoid profile was distinctly modulated, with certain compounds being significantly upregulated or downregulated in specific comparisons.

To further visualize the specific terpenoids (ursane-type triterpenes) that were most significantly altered in key varietal comparisons, we performed differential abundance heatmaps (Figure 7). In the comparison between the high-terpenoid DY and the wild type (WL), nearly all analyzed ursane-type triterpenes were obviously abundant in DY, with compounds like 3β,6β,19α,24-tetrahydroxyurs-12-en-28-oic acid and tormentic acid showing particularly strong upregulation (Figure 7A). This pattern reinforces DY’s unique phytochemical profile centered on these compounds. The contrast between the cultivated GC and SG (both belong to A. chinensis) genotype showed that the SG variety consistently displayed higher levels of most ursane-type triterpenes, including ursolic acid and corosolic acid, whereas its cultivated counterpart GC showed a general suppression of this pathway (Figure 7B). A similar trend was observed when comparing the two A. arguta cultivars, LC and DY (Figure 7C), where DY again demonstrated a higher abundance of these triterpenoid acids. Moreover, the comparison between LC and SG highlighted SG’s unique ability to accumulate specific ursane-type triterpenes than LC (Figure 7D). Collectively, these targeted heatmaps crystallize the identification of ursane-type triterpenes as key metabolic differentiators, with the DY and SG varieties representing potent genetic resources for these compounds.

To evaluate the fold change differences of terpenoids between genotypes, a pairwise comparative analysis was conducted. The results revealed that DY variety exhibited a pronounced upregulation of triterpenoid acids compared to GC, WL, and SG (Table 1). Key markers such as 1-oxo-siaresinolic acid, camellisin B, and siegesbeckic acid were consistently elevated by over 400-fold (Log₂FC > 8.7) in these comparisons. Conversely, DY showed an obvious downregulation of seco-iridoid glucosides like schizonepetoside A and geniposide when compared to SG and GC, with Log₂FC -8.79 (Table 1). The LC variety also displayed a significant upregulation of diverse terpenoids such as dehydroisoactinidic acid (294-fold up) and dihydrocatalpol (171-fold up) compared to GC and SG. However, like DY, LC was characterized by very low levels of certain glucosylated terpenoids, including blumenol C glucoside (Log₂FC -7.1) and vogeloside (Log₂FC -4.84) as compared to SG. The SG was uniquely enriched in a range of compounds including geniposide (442-fold up) and several apocarotenoids (Log₂FC > 7) (such as 3-Hydroxy-13-Apo-ϵ-Caroten-13-one, blumenol C glucoside; byzantionoside B) and phenolic-terpenoid conjugates like 3’-O-Caffeoyl sweroside Log₂FC > 6.7, which were nearly absent or much lower in WL (Table 1). This suggests that the SG variety possesses a specialized metabolic profile distinct from the wild type.

The antioxidant capacity significantly varied among the five kiwifruit varieties (Figure 8). The SG exhibited the highest mean antioxidant capacity (455.20 ± 3.12 µg TE/g FW), followed by the Danyang (DY) variety (438.91 ± 3.43 µg TE/g FW) (Figure 8). The LC, GC, and WL fruits showed statistically similar, but lower, antioxidant capacity, with values of 425.63, 419.60, and 415.70 µg TE/g FW, respectively. These results indicate that the SG and DY kiwifruit varieties possess a superior in vitro antioxidant potential, which may contribute to their enhanced nutritional value and shelf-life.