Plant material and treatments

The experiment was conducted at the National Apple Industry Technology System Pingliang Apple Comprehensive Experiment Station, located in Jingning Demonstration County, Chengchuan Town, China (35°24′16″N, 105°47′15″E) in 2022. The experimental trees were 15-year-old ‘Changfu No. 2’ Fuji apple trees grafted onto Malus robusta rootstocks. The trees were planted at a spacing of 4 m × 3 m, leading to a density of 834 trees per hectare. The trees were trained in an improved spindle shape with heights ranging between 4.2 and 4.6 m, a trunk height of 60 to 75 cm, and six to nine main branches angled at between 70° and 80°.

The thinning and reshaping restructuring experiment was initiated in mid-February 2022 in the experimental orchard. A completely randomized block design with thinning and restructuring as the single experimental factor, consisting of two treatments: T (thinning with restructuring) and CK (control without thinning or restructuring), was adopted. In the T treatment, alternate trees were removed to adjust the original planting density from 4 m × 3 m to 4 m × 6 m spacing. A comprehensive restructuring of the canopy was achieved through the selective removal of large branches, elevation of trunk height to 100–110 cm, and elimination of the central leader, resulting in tree heights of 310–340 cm. Four to six healthy, well-distributed scaffold branches were retained as permanent structural elements with branch angles systematically adjusted to 90°–100°. Standard Fuji apple pruning protocols were followed to achieve a total pruning intensity of approximately 30%. The control group (CK) maintained the original planting density and canopy architecture throughout the experiment. Each 500 m² treatment plot contained three replicates with five representative trees of moderate vigor. There were no visible pests or disease symptoms in any of the replicates. Sampling was conducted during critical phenological stages on June 22 (early physiological differentiation) and August 22, 2022 (late physiological differentiation). Sampling entailed collecting terminal buds from medium- and short-shoots (length < 15.0 cm) in the mid-lower canopy region, with strict adherence to uniformity criteria for bud size and developmental stage. Samples were individually wrapped in pre-labeled aluminum foil immediately after collection, flash-frozen in liquid nitrogen, and stored at −80 °C to preserve biochemical integrity for subsequent analysis.

Determination of carbohydrate content

The contents of sucrose, glucose, fructose, and sorbitol in the treatment and control groups were measured using the respective commercially available assay kits. These kits were purchased from Merck KGaA (Darmstadt, Germany) with the following product reference numbers: SCA20 for the sucrose assay kit, MAK013 for the glucose assay kit, FA20 for the fructose assay kit, and PHR1006 for the sorbitol assay kit.

Metabolomics analysis

Apple blossom buds (100 mg) were homogenized in 300 µL of 50% methanol and incubated at 4 °C overnight. The mixture was then centrifuged at 13,000 rpm for 20 min at the same temperature. The supernatant was then carefully collected and concentrated using a vacuum freeze dryer. The dried residue was diluted in 100 µL of methanol for LC/MS analysis. A Nexera X2 system (Shimadzu, Kyoto, Japan) was coupled with an AB Sciex Triple TOF 5600 system (AB Sciex, Framingham, MA, USA) for LC/MS analysis, resulting in a substantial amount of mass spectrometry data.

Non-targeted metabolomic analysis

Frozen samples (−80 °C) were thawed on ice, homogenized with 400 µL of methanol/water (7:3, v/v) containing an internal standard (20 mg sample), and vortexed for 3 min. The homogenate was subsequently sonicated in ice for 10 min, vortexed for 1 min, and incubated at −20 ° C for 30 min. The mixture was then centrifuged (12,000 rpm, 10 min, 4 °C) to obtain the supernatant. The supernatant was re-centrifuged (12,000 rpm, 3 min, 4 °C), and 200 µL of the clarified supernatant was subjected to LC-MS analysis. The settings of the positive ion mode were: separation on a Waters ACQUITY Premier HSS T3 column (1.8 µm, 2.1 × 100 mm) using 0.1% formic acid in water (A) and 0.1% formic acid in acetonitrile (B) with the gradient set as follows: 5% B → 20% B (2 min), → 60% B (3 min), → 99% B (1 min), hold 1.5 min, → 5% B (0.1 min), and equilibrate for 2.4 min. The column temperature was 40 °C, the flow rate was 0.4 mL/min, and the injection volume was 4 µL. The settings of the negative ion mode were similar to those of the positive-ion mode except for ionization polarity.

Data processing and bioinformatics analysis

Signal intensity information for each compound across different samples was extracted using the XCMS software and processed using the metaX software for quality control. The workflow involved removing low-quality peaks (those missing in more than 50% of QC samples or missing in over 80% of actual samples), imputing missing values using the KNN (R (impute) software, Version 1.56.0, default parameters), and data normalization through probabilistic quotient normalization (PQN) followed by QC-robust spline batch correction (QC-RSC). The corrected and filtered peaks were subjected to metabolite identification by searching against the laboratory’s in-house database and integrating public databases (HMDB, https://hmdb.ca/, and KEGG, https://www.kegg.jp/). The annotated metabolites were then mapped to the KEGG Pathway database (http://www.kegg.jp/kegg/pathway.html). Unsupervised principal component analysis (PCA) was conducted using the prcomp function in R (http://www.r-project.org) for dimensionality reduction and pattern recognition. Orthogonal partial least squares-discriminant analysis (OPLS-DA) was conducted using the R package MetaboAnalystR to extract variable importance in projection (VIP) values and generate model diagnostics. Differential metabolites between groups were identified using the following criteria: VIP > 1 (from OPLS-DA), fold-change thresholds (≥2 or ≤0.5), and statistical significance of P < 0.05 (two-tailed Student’s t-test).

Phenotypic and physiological characteristics

The differentiation of apple flower buds was examined following thinning and reshaping treatment. The flower buds were gathered and photographed at the late stage of flower bud differentiation (T2 vs CK2). Flower buds in the T2 group were significantly bigger and fuller than those in the CK2 group. They also had more layers in their longitudinal section (Figs. 1A–1B). The levels of carbohydrates (such as sucrose, glucose, fructose, and sorbitol, among others) varied significantly throughout the flower bud differentiation process. The sucrose, sorbitol, fructose, and glucose levels in T1 were significantly higher than those in CK1 during the early stages of flower bud development. Similarly, T2 had significantly higher amounts of sucrose, sorbitol, fructose, and glucose than CK2 during the late stage of flower bud differentiation. These findings suggested that thinning and reshaping positively promote flower bud differentiation during the early and late stages (Figs. 1C–1F).

Multivariate statistical analysis

Each group had six biological replicates with high consistency to ensure data reliability. Quality control analyses were conducted before the study to ensure the reliability of the detection results and good intra-treatment replicability (Fig. S1). PCA results revealed significant separation of T1 and CK1 samples (Fig. 2A) and T2 and CK2 samples (Fig. 2B), indicating significant differences between treatments. These results highlighted the significant metabolic profile differences between T1 and CK1, and between T2 and CK2.

The TIC maps were quite identical, which demonstrated the instrument’s strong stability and data accuracy (Fig. S1).

Metabolic profiling

Metabolomic composition analysis revealed hierarchical characteristics in the metabolite distribution among different experimental groups. Secondary classification of 1,312 metabolites from the T1 vs CK1 comparison group revealed the following top five metabolite categories and their proportions: lipids and lipid-like molecules (23.3%), phenylpropanoids and polyketides (20.6%), organic acids and derivatives (10.8%), benzene-ring-containing compounds (10.6%), and heterocyclic compounds (9.4%) (Table S1A). The specific compositional feature of dominant categories in the lipids and lipid-like molecules category was fatty acyls (39.7%) (Table S1B). The phenylpropanoids and polyketides category was dominated by flavonoids (48.1%) and cinnamic acids and derivatives (20.0%) (Table S1C). The organic acids and derivatives category was dominated by carboxylic acids and derivatives (80.9%) (Table S1D). The benzene-ring-containing compounds category was dominated by benzene and substituted derivatives (74.8%) (Table S1E). The heterocyclic compounds category was dominated by imidazopyrimidines (10.6%) and indoles and derivatives (9.8%) (Table S1F). Notably, the top five secondary classifications in the T2 vs CK2 comparison group involving 1,306 metabolites remained completely consistent with those of the T1/CK1 group (Table S2A). However, there were distinct differences in subclass distributions. The lipids and lipid-like molecules category was dominated by fatty acyls (39.7%) and terpenoid lipids (30.5%) (Table S2B). The subclass compositions of the other four major categories (Tables S2C–S2F) remained highly consistent with those of the T1/CK1 group, with no significant changes in the proportions of the dominant subclasses, i.e., flavonoids (48.1%), carboxylic acid derivatives (80.9%), benzene-substituted compounds (74.8%), and imidazopyrimidines (10.6%).

Composition differences of the metabolites

There were 60 metabolites with significantly different expressions (32 up-regulated and 28 down-regulated) in the T1 vs CK1 comparison group (Fig. 3A). In contrast, the T2 vs CK2 comparison group had 51 metabolites with significantly different expressions (26 up-regulated and 25 down-regulated) (Fig. 3B). The top 10 metabolites with the largest absolute value of Log2FC in each comparison group were selected based on the screening criteria for the drawing of the radar chart. The top 10 metabolites in the T1 vs CK1 comparison group based on difference multiple were trifloxystrobin, difenoconazole, pyraclostrobin, propiconazole, 11(12)-Epoxy-5Z,8Z,14Z-eicosatrienoic acid, acetamiprid, hippuric acid, methyl ester, coumarin, 4-Formylbenzonitrile, and spirodiclofen (Fig. 3C). In the T2 vs CK2 comparison group, the top 10 metabolites based on difference multiple were trifloxystrobin, difenoconazole, pyraclostrobin, propiconazole, prodione, flusilazole, tebuconazole, m-Hydroxybenzoylecgonine, acetamiprid, and 4-formylbenzonitrile (Fig.  3D).

KEGG metabolic pathway analysis and screening of metabolites related to flower bud differentiation

KEGG metabolic pathway enrichment analysis (p < 0.05, fold change (FC) ≥2 or FC≤0.5) was done to better understand the mechanism by which thinning and reshaping enhance apple flower bud differentiation.

Notably, biosynthesis of secondary metabolites was the most significant common metabolic pathway in the two comparison groups. Biosynthesis of amino acids of secondary metabolites, flavonoids, flavone and flavonols, biosynthesis of various plant secondary metabolites, phenylpropanoid biosynthesis, and biosynthesis of various alkaloids were significantly enriched in the T1 vs CK1 comparison group (Fig. 4A). Eight metabolites were enriched in these pathways. Biosynthesis of amino acids of secondary metabolites, flavonoids, flavone and flavonols, biosynthesis of various plant secondary metabolites, phenylpropanoid biosynthesis, biosynthesis of various alkaloids, alpha-linolenic acid metabolism, plant hormone signal transduction, pentose phosphate pathway, and carbon metabolism were all significantly enriched in the T2 vs CK2 comparison group. A total of 10 metabolites were enriched in these pathways (Fig. 4B).

Histidine (C00135) participated in the biosynthesis of amino acids, and eriodictyol (C05631) participated in flavonoid biosynthesis. In contrast, coumarin (C05851) participated in the biosynthesis of various plant secondary metabolites in the two comparison groups (Tables S3A & S3B).

In the T1 vs CK1 comparison group (Table S3A), 2-isopropylmalic acid (C02504) and 4-hydroxyphenylpyruvic acid (C01179) were involved in the biosynthesis of various plant secondary metabolites. (+)-catechin (C09727) was involved in flavonoid biosynthesis, 3′, 5′-dimethoxy-3, 5, 7, 4’-tetrahydroxyflavone (C11620) in flavone and flavonol biosynthesis, and trans-ferulic acid (C01494) in phenylpropanoid biosynthesis and biosynthesis of various alkaloids.

Umbelliferone (C09315) and isopimpinellin (C02162) participated in the biosynthesis of various plant secondary metabolites in the T2 vs CK2 comparison group (Table S3B), which was also included. Myricetin (C10107) was involved in flavonoid and flavone and flavonol biosynthesis. Trans-caffeic acid (C01197) was involved in phenylpropanoid biosynthesis and biosynthesis of various alkaloids. Jasmonic acid (C08491) and transtraumatic acid (C16308) were involved in alpha-linolenic acid metabolism. Tasmonic acid (C08491) was involved in plant hormone signal transduction. 6-phosphogluconic acid (C00345) was involved in the pentose phosphate pathway and carbon metabolism.

L-histidine, 2-Isopropylmalic acid, eriodictyol, (+)-catechin, coumarin, and trans-Ferulic acid were relatively increased in the T1 vs CK1 comparison group (Fig. 5). In contrast, 3′, 5′-Dimethoxy-3, 5, 7, 4′-tetrahydroxyflavone and 4-hydroxyphenylpyruvic acid were relatively reduced in this comparison group. Isopimpinellin, coumarin, trans-caffeic acid, jasmonic acid, eriodictyol, and 6-phosphogluconic acid were relatively increased in the T2 vs CK2 comparison group (Fig. 6). In contrast, myricetin, histidine, umbelliferone, and trans-traumatic acid were relatively reduced in this comparison group. Notably, eriodictyol and coumarin exhibited a consistent increase in both comparison groups, indicating that their functions were potentially similar in the process of regulating flower bud differentiation. Histidine exhibited an opposite trend in the two comparison groups, suggesting that its functions potentially change in the process of regulating flower bud differentiation.

Flower bud differentiation is closely associated with carbohydrate content

Carbohydrates are essential structural compounds and energy sources in plant growth and development. They play a crucial role in flower bud differentiation. Thinning and reshaping of fruit orchards enhance light intensity, thereby enhancing flowering. Studies postulate that carbohydrates are involved in the flowering process of chrysanthemums under photoperiod induction (Wang et al., 2010). Glucose and fructose are directly utilized to provide sufficient nutrients for flower bud differentiation. Notably, their content is positively correlated with the differentiation process. Sorbitol may also be physiologically transformed into fructose and glucose (Morandi et al., 2008), thereby promoting flower bud differentiation. Sucrose is the primary product of photosynthesis and is the main form of sugar transport in fruit trees. It also acts as a signaling factor for flowering in plants (Usenik, Fabčič & Štampar, 2008). Xing et al. (2015) demonstrated that the glucose and fructose content in the terminal flower buds and adjacent leaves of ‘Nagafu 2’ apples change significantly during the flower induction stage. In the same line, Barbier et al. (2015) postulated that sucrose accumulation promotes flower bud differentiation in roses. A study by Xu et al. (2023) reported that the sorbitol content significantly increases in loquat flower buds during differentiation, with exogenous sorbitol application further promoting flower bud formation. Herein, thinning and reshaping of the apple orchard increased light intensity, thereby increasing the amount of glucose, fructose, sucrose, and sorbitol in both the early and late phases of flower bud differentiation and accelerating the differentiation process.

Flower bud differentiation is closely associated with secondary metabolites

Plant hormones influence plant germination, flowering, fruiting, dormancy, abscission, and participate in floral regulation (Du, Yu & Zhou, 2021; Milyaev et al., 2022). Different phytohormones promote or inhibit flower bud differentiation. Plants transition from vegetative growth to reproductive growth by regulating the balance of various hormones (Iqbal et al., 2017). Previous studies postulate that endogenous tZR content increases progressively during flowering (Pang et al., 2014). Disrupted gibberellin biosynthesis and signal transduction lead to delayed flowering in GA mutant plants (Lin et al., 2012), while exogenous BR application delays flowering in wild-type Arabidopsis (Noh et al., 2004). Jasmonic acid (JA) is a lipid-derived phytohormone that participates in root elongation, petal growth, stamen development, senescence promotion, and stress response (Sheng, 2020). JA enhances GA signaling pathways to promote floral organ formation (Hou et al., 2010). Endogenous hormones such as IAA, JA, and SA positively regulate bud germination and floral morphogenesis in kiwi fruit. In contrast, elevated tZR and ABA levels combined with reduced GA1 and SAG content facilitate leaf-to-flower bud transition (Zhang, 2024). Exogenous SA application induces endogenous SA synthesis and suppresses the expression of transcription factor FLC in wild-type Arabidopsis, thereby promoting flowering (Khurana & Cleland, 1992). Herein, the T2 group exhibited larger and fuller flower buds with more layered structures in longitudinal sections compared to the CK2 group. The increased JA content suggested its potential role in promoting flower bud differentiation.

The phenylpropanoid biosynthesis (map00940) pathway is crucial for flower bud differentiation (Yao et al., 2021). In this study, it was associated with biosynthesis of various plant secondary metabolites (map00999), flavonoid biosynthesis (map00941), and biosynthesis of various alkaloids (map00996). Biosynthesis of various plant secondary metabolites (map00999) included coumarin biosynthesis. Coumarins are crucial in keeping the structural integrity and signal transduction during flower bud differentiation (Fan et al., 2018). Notably, both T1 vs CK1 and T2 vs CK2 comparison groups exhibited parallel increases in coumarin and eriodictyol concentrations, suggesting potential functional synergy between these metabolites in bud differentiation. The T1 group exhibited elevated levels of 2-isopropylmalic acid, (+)-catechin, and trans-ferulic acid, indicative of their positive regulatory roles. Conversely, 3′, 5′-dimethoxy-3,5,7,4′-tetrahydroxyflavone and 4-hydroxyphenylpyruvic acid exhibited a decrease in concentration in these comparisons. In the T2 vs CK2 group, myricetin and isopimpinellin demonstrated significant upregulation, while umbelliferone and trans-caffeic acid exhibited inverse concentration trends. These differential accumulation patterns highlight the temporal specificity of metabolic regulation during flower bud differentiation.

Biosynthesis of various alkaloids (map00996) was associated with plant hormone signal transduction (map04075) and alpha-linolenic acid metabolism (map00592). Biosynthesis of various plant secondary metabolites (map00999) was associated with citrate cycle (map00020), carbon metabolism (map01200), which contained pentose phosphate pathway (map00030), and citrate cycle (map00020). The pentose phosphate pathway was associated with histidine metabolism (map00340). Notably, jasmonic acid and 6-phosphogluconic acid were significantly increased in T2 vs CK2. In contrast, the comparison group exhibited a decrease in trans-traumatic acid levels. These findings collectively reveal dynamic metabolic reprogramming events during bud differentiation.

The observed metabolic shifts and pathway-specific activation/inhibition signatures suggested a complex regulatory network of secondary metabolism to drive floral developmental transitions.