The Early-Mid Ripe mango cultivar “Golden Phoenix” was planted in Yuanjiang Hani, Yi, and Dai Autonomous County (101°39′-102°22′N, 23°18′-23°55′W) of Yuxi City, Yunnan Province, China. Mango materials were provided by Yuanjiang County Agricultural Technology Extension Service Center in Yunnan province. Experimental research on mangoes was conducted in compliance with China laws and regulations and local legislations in Yunnan province. The mango trees were 8 years old with a row spacing and plant spacing of 3 × 5 m, respectively. MG fruits were harvested 120 days after flowering (reaching physiological maturity), and TR fruit was harvested about 150 days after flowering in the summer of 2021. According to the project listed by Supplementary Table S1, their fruit quality was measured as follows: transverse and longitudinal diameters of the fruit were measured with a Vernier caliper, and the fruit shape index was the ratio of the longitudinal to transverse diameters. Fruit texture was measured using a durometer (Fujiwara FHM-1, Japan). Fruit total soluble solids content was measured with a hand-held refractometer (ATAGO PAL-1, Japan). Titratable acid content was determined by NaOH titration. Excel 2016 was used for data sorting and Origin 8.5 for chart drawing. Correlation analysis was performed using SPSS 19.0 (SPSS Inc., Chicago, IL, United States). Data from all analyses were expressed as average and standard error. The threshold of significance was set at p < 0.05. Each group consisted of 20 fruits, in triplicate, and samples were designated MG 1, 2, 3 and TR 1, 2, 3, respectively. Three biological replicates were set for each sample, and 20 fruits were collected randomly from five trees of the same growth in each replicate. After taking mangos back to the laboratory, the middle part of pulps (about 15 g in weight) was carefully excised with a razor blade, immediately frozen in liquid nitrogen, and stored at-80°C for further analysis.

Mango fruit samples were well ground in liquid nitrogen and crushed at 30 Hz for 1.5 min using a hybrid mill with zirconia beads (MM400, Retsch). Powdered samples (100 mg) were weighed and extracted overnight at 4°C with 1.0 ml of 70% aqueous methanol. Following centrifugation at 10,000 rpm for 10 min, the extracts were filtrated (SCAA-104, 0.22 μm pore size; ANPEL, Shanghai, China)¹ for UPLC–MS/MS analysis. The metabolite extraction, identification, and quantification were performed previously (Wang et al., 2017a).

The sample extracts were analyzed using an UPLC-ESI-MS/MS system (UPLC, SHIMADZU Nexera X2,² MS, Applied Biosystems 4,500 Q TRAP).³ The analytical conditions were as follows, UPLC: column, Agilent SB-C18 (1.8 μm, 2.1 mm*100 mm); the mobile phase was consisted of solvent A, pure water with 0.1% formic acid, and solvent B, acetonitrile with 0.1% formic acid. Sample measurements were performed with a gradient program that employed the starting conditions of 95% A, 5% B. Within 9 min, a linear gradient to 5% A, 95% B was programmed, and a composition of 5% A, 95% B was kept for 1 min. Subsequently, a composition of 95% A, 5.0% B was adjusted within 1.10 min and kept for 2.9 min. The column oven was set to 40°C; the injection volume was 4 μl. The effluent was alternatively connected to an ESI-triple quadrupole-linear ion trap (Q TRAP)-MS (AB SCIEX, United States). Triple quadrupole-linear ion trap mass spectrometer (QTRAP; API 4500 Q TRAP LC/MS/MS System) was used for Linear Ion Trap (LIT) and triple quadrupole (QQQ) scans. QQQ and LIT modes with 10 and 100 μmol/l polypropylene glycol solutions, respectively. Metabolite data analysis and quantification were performed using Analyst 1.6.1 software (AB SCIEX, Ontario, Canada) and multiple reaction monitoring (MRM), respectively. Finally, the identified metabolites were subjected to partial least squares discriminant analysis (PLS-DA). Significantly regulated metabolites between groups were determined by VIP ≥ 1, value of p < 0.05, and absolute Log2FC (fold change) ≥ 1. VIP values were extracted from OPLS-DA result, which also contain score plots and permutation plots, was generated using R package MetaboAnalystR. The data were log transform (log2) and mean centering before OPLS-DA. In order to avoid overfitting, a permutation test (200 permutations) was performed (Zhang et al., 2020).

Mango samples were ground into a powder in liquid nitrogen. 1 g of the mango powder was weighed and transferred immediately to a 20 ml headspace vial (Agilent, Palo Alto, CA, United States), containing NaCl saturated solution to inhibit any enzyme reaction. The vial was sealed using a crimp-top cap with TFE-silicone headspace septa (Agilent). At the time of solid-phase microextraction (SPME) analysis, the vial was placed in 100°C for 5 min; then, a 120 μm divinylbenzene/carboxen/polydimethylsiloxane fiber (Agilent) was exposed to the headspace of the sample for 15 min at 100°C. After sampling, desorption of the VOCs from the fiber coating was carried out in the injection port of the GC apparatus (Model 8,890; Agilent) at 250°C for 5 min in the splitless mode. The identification and quantification of VOCs were carried out using an Agilent Model 8,890 GC and a 5977B mass spectrometer (Agilent), equipped with a 30 m × 0.25 mm × 0.25 μm DB-5MS (5% phenyl-polymethylsiloxane) capillary column. Helium was used as the carrier gas at a linear velocity of 1.2 ml/min. The injector temperature was kept at 250°C and the detector temperature at 280°C. The oven temperature was programmed from 40°C (3.5 min), increasing at 10°C/min to 100°C, at 7°C/min to 180°C, and at 25°C/min to 280°C, hold for 5 min. Mass spectra were recorded in electron impact (EI) ionization mode at 70 eV. The quadrupole mass detector, ion source, and transfer line temperatures were set, respectively, at 150, 230 and 280°C. Mass spectra were scanned in the range m/z 50–500 amu at 1 s intervals. Identification of volatile compounds was achieved by comparing the mass spectra with the data system library (MWGC or NIST) and linear retention index (Chai et al., 2012; Guo et al., 2016; Xia et al., 2021).

The mango pulp was freeze-dried by liquid nitrogen and ground into a powder in a mortar. 50 mg of the dried powder was weighed and put into a mixture of hexane, acetone, and ethanol, followed by addition of the internal standard. The stock solution of the standard was stored at a concentration of 1 mg/ml at-20°C. Next, the extract was vortexed for 20 min at room temperature and centrifuged (10,000 rpm for 10 min, 5424R, Eppendorf, German), and the supernatant was collected, then evaporated to dryness under a stream of nitrogen, and reconstituted in a mixture of methanol and methyl tertiary butyl ether (MTBE). The mixed solution was then filtered through a 0.22 μm filter for further LC–MS analysis. Subsequently, LC- Atmospheric Pressure Chemical Ionization (APCI)-MS/MS (UHPLC, ExionLC™ AD,⁴ MS, Applied Biosystems 6,500 Triple Quadrupole; see footnote 4) was conducted under the following conditions: HPLC column: YMC C30 (3 μm, 100 mm × 2.0 mm i.d), solvent system consisting of methanol, acetonitrile (1: 3, v/v) with 0.01% butylated hydroxytoluene (BHT) and 0.1% formic acid (A), and methyl tert-butyl ether with 0.01% BHT (B), gradient program: 0% B (0–3 min), 70% B (3–5 min), 95% B (5–9 min), and 0% B (11–12 min), flow rate: 0.8 ml/min, temperature: 28°C, and injection volume: 2 μl. A API 6500 + Q TRAP LC/MS/MS system, equipped with an APCI Turbo Ion-Spray interface, was utilized in a positive ion mode and controlled by Analyst 1.6.3 software (AB Sciex). The APCI source operation parameters included ion source: APCI+, source temperature: 350°C, and curtain gas (CUR): 25.0 psi. Following optimization, DP and CE were conducted for individual MRM transition. In the end, a specific set of MRM transitions were monitored during each period according to the carotenoids eluted within this period (Rodrigues et al., 2016; Petry and Mercadante, 2018).

A total of six transcriptome sequencing libraries were constructed using two mango pulp samples with three biological replicates each. Total RNA extraction from mango pulps was performed using the TRIzol method (Xia et al., 2021). RNA concentration and purity were determined by NanoDrop 2000 (NanoDrop Technologies, Wilmington, DE, United States) and Agilent Bioanalyzer 2,100 system (Agilent Technologies, Palo Alto, CA, United States), respectively. The integrity of the RNA was determined using 1% agarose gel electrophoresis. Afterward, mRNA was isolated from total RNA using magnetic beads with oligo (dT), and complementary deoxyribonucleic acid (cDNA) was synthesized by ligating sequencing adapters to both ends using a cDNA synthesis kit (TaKaRa, Japan). Next, library preparations were sequenced on the Illumina HiSeq 4,000 platform, and single gene sequences obtained from the Green Plant Transcriptome database were integrated and annotated by RSEM 1.3.1 software (Dewey and Li, 2011).

Unigene sequences were compared with KEGG, NR, Swiss-Prot, GO, COG/KOG, TrEMBL databases using BLAST software, and the amino acid sequences of unigenes were predicted and compared with Pfam database using HMMER software to obtain annotation information of unigenes. The unnormalized reads count data of the gene were entered on DESeq2. Next, the Benjamini-Hochberg method was used to correct multiple hypothesis testing probabilities (value of p). Differential genes were screened based on |Log₂Fold Change| ≥ 2 and FDR < 0.05. The fragments per kilobase of exon model per million mapped reads (FPKM) of genes were centralized and normalized, and then, Kmeans clustering analysis was performed. Subsequently, after annotating the genes into the KEGG database, the number of differential genes contained in each KEGG pathway was counted, and the significantly enriched pathway was identified among the differentially expressed genes. Finally, the distribution of differential genes in the Gene Ontology (GO) was investigated using enrichment analysis to elucidate the functional representation of sample differences in genes in the experiment. Plant TF prediction was performed using iTAK 1.7 software,⁶ which integrates two databases, namely, PlnTFDB and PlantTFDB (Zheng et al., 2016).

Total RNA was extracted from mango fruits by the CTAB method (Wang et al., 2019). Based on the transcriptome data of TR vs. MG mango fruits, the expression level of 13 carotenoid-biosynthetic pathway genes and seven TFs (MYB, bHLH, and NAC) was validated. The PCR was performed with an ABI 7500 Fast Real-Time Detection System (Applied Biosystems, Waltham, MA, United States) using the Ultra SYBR Mix Kit (TaKaRa, Dalian, China). In the amplification system (the total volume of 20.0 μl), where there were 10.0 μl of Ultra SYBR Premix System II, 0.5 μl of 10 μM upstream primer, 0.5 μl of 10 μM down-stream primer, 2.0 μl template, and 7.0 μl double-distilled water, the amplification was conducted at 95°C for 8 min, followed by 38 cycles of 95°C for 5 s and 60°C for 30 s. Relative quantitative analysis of data was performed by the 2^−ΔΔCT method with β-actin as the reference gene. The primers used for RT-qPCR are listed in Supplementary Table S2.

Mangoes of the cv. “Golden Phoenix” at fruit ripening stages of MG and TR are shown in Figure 1A. MG and TR stages mangoes were sequenced with three biological replicates for each stage. The paired-end reads of cDNA libraries of each were obtained using the Illumina HiSeq 4,000 platform and are shown in Supplementary Table S3. The mapping rates obtained though comparison with the reference database were 84.80 and 89.03% (Supplementary Table S3). Differentially expressed genes (DEGs) identified among the two samples were filtered for an FDR < 0.05 and a Log₂FC ≥ 2. The data indicated 20,120 DEGs between MG and TR, with the number of upregulated genes slightly larger than that of downregulated genes (12,870 vs. 7,250; Figure 1B). Biological pathways were assigned to the DEGs from the KEGG database after screening for pathways with a value of p < 0.05. The results show that the DEGs in “TR vs. MG” were significantly enriched in plant hormone signal transduction, carbon metabolism, pyruvate metabolism, metabolic pathways, and biosynthesis of secondary metabolites (Figure 1C). Substantial and significant enrichment was observed in both metabolic pathways and biosynthesis of secondary metabolites. These metabolic pathways shed light on the metabolic processes occurring during the development of mango fruits. From the GO⁷ annotations of the DEGs, it can be seen that 1,154 unigenes were annotated as “Biological Process,” 674 as “Cellular Component,” and 572 as “Molecular Function” (Figure 1D).

To understand the molecular mechanism leading to differences in flavor compounds in the development of “Golden Phoenix” mangoes, a metabolomic comparison of MG and TR stages using high performance liquid chromatography-mass spectrometry (HPLC-MS/MS) was conducted. Partial least squares discriminant analysis (PLS-DA) of the HPLC-MS/MS data showed a clear difference between MG and TR samples along PC1 which accounts for 87.1% of the total variance (Figure 2A). The majority of the residual variance (PC2; 7.76%) appears to be due to differences in biological replicates. Using metabolite concentration data for the cluster analysis of a stratified heat map of the samples, it was observed that all biological replicates were grouped together (top of the figure), which indicates a high reliability of the resulting metabolome data (Figure 2B). The metabolomic analysis revealed that 309 metabolites showed differential levels in MG vs.TR fruits, including organic acids, esters, terpenes, saccharides, and alcohols (Supplementary Table S4). After filtration of these metabolites for those with a Log2FC ≥ 1 or ≤ −1, and a variable importance projection (VIP) ≥ 1, 214 upregulated and 95 downregulated metabolites were identified in TR vs. MG (Figure 2C). The pathways associated with the metabolites thus selected were identified by use of the KEGG database. The results indicated these metabolites were mainly enriched in pathways including metabolic pathways, citrate cycle (TCA cycle), biosynthesis of secondary metabolites, and biosynthesis of amino acids (Figure 2D).

The software package⁸ “Cor” in the language package R was used to calculate the Pearson correlation coefficient (PCC) of DEGs and differential metabolites, with the fold-changes of metabolites with a PCC > 0.8 in each group shown in the nine-quadrant diagram (Supplementary Figure S1). In the third and seventh quadrants, a total of 12,839 DEGs and 71 metabolites were detected in “TR vs. MG.” The consistence in differences in gene expressions and metabolites indicates that in general, the changes in gene expression drive the observed changes in metabolite levels. A two-way orthogonal PLS (O2PLS) model was established for all DEGs and differential metabolites and according to the load chart, variables with high relevance and significant weight in the different datasets were preliminarily evaluated to select important metabolites (Supplementary Figure S2). The nine metabolites most affected by the transcriptome included 2 amino acids and derivatives (Benzyl-(2”-O-xylosyl)glucoside; N-acetyl-L-leucine), 2 nucleotides and derivatives (9-(arabinosyl)hypoxanthine; Guanine), 3 phenolic acids (4-O-glucosyl-sinapate; 2-O-Trigalloyl-glucose–glucose; and Dicaffeoylshikimic acid), and 2 lipids (LysoPC 18:2; LysoPC 19:0) and three of these metabolites (N-acetyl-L-leucine; Dicaffeoylshikimic acid; and 2-O-trigalloyl-glucose–glucose) presented significant differences between MG and TR stages.

Major contributors to the flavor and nutritional content of fruit include organic acids, amino acids, and saccharides. With a VIP ≥ 1, value of p < 0.05, and |Log₂FC| ≥ 1 as the threshold values for significant difference, a total of 115 flavor-related metabolites with significantly differential expressions in MG and TR were identified, including 45 amino acids and derivatives, 50 organic acids, and 20 saccharides and alcohols (Supplementary Table S4). Moreover, most flavor compounds showed a trend of increase in TR mango fruit and we identified six metabolites among 45 amino acids and derivatives metabolites with a Log2FC ≥ 4, of which four were upregulated and two were downregulated (Table 1). Among the metabolites showing significant increases, Glutathione (GSH), O-phospho-L-serine, γ-glu-cys, and Hexanoyl-L-glycin were increased with fold-changes of 18.78, 12.62, 11.28, and 10.98 expression, respectively. The organic acids (and derivatives), 2-hydroxy-2-methylbutyric acid, tartronate semialdehyde, 4, 5, 6-Trihydroxy-2-oxohexanoic acid, phosphoen olpyruvate, 6-hydroxyhexanoic acid, malonic acid, and DL-glyceraldehyde-3-phosphate were upregulated by 10.13, 11.58, 12.67, 14.58, 14.73, 15.32, and 15.92 times, respectively. Three saccharides and alcohols (3-phospho-D-glyceric acid, D-glucuronic acid, and D-fructose-1,6-biphosphate) were increased with fold-changes of 19.73, 16.39, and 14.78, respectively (Table 1). These differential metabolites may play a key role in the formation of TR fruit taste during fruit ripening.

Plant VOCs are secondary metabolites that play an important role in the flavor of fruits (Coelho et al., 2010). In mango fruits, VOCs mainly consist of esters, alcohols, terpenoids, heterocyclic compounds, aldehydes, and acids. PCA revealed a clear separation of volatile metabolite profiles of the pulps from two varieties, with PC1 and PC2 representing 85.33% of the total components (Figure 3A). With VIP ≥ 1, value of p < 0.05, and |Log2FC| ≥ 1 as the threshold values of significant difference, a total of 82 volatile metabolites with significantly differential expressions were identified in “TR vs. MG” (Figure 3B; Supplementary Table S5). Of these, 28 esters and 22 terpenoids act as the main aroma metabolites and constitute the unique aroma of the mango fruit. To screen more precisely for the major volatile metabolites in mango fruit, we narrowed the screening range to value of p < 0.01 and |Log2FC| ≥ 4 as thresholds and showed that among the 12 downregulated metabolites, three terpenoids (Naphthalene-octahydro-dimethyl-7-naphthalene, 1,3-dimethyl-5-(propen-1-yl) adamantane, and dimethyl-7-decahydronaphthalen-1-ol) showed more than 10-fold decrease in TR compared to that in MG, Log2FC was −11.15, −11.27, and − 11.54 in expression. In addition, esters occupied six of the seven upregulated metabolites, including octanoic acid, ethyl ester, butanoic acid, butyl ester, butanoic acid, hexyl ester, ethyl tridecanoate, butanoic acid, 2-methylpropyl ester, and propanoic acid, 2-methyl- and octyl ester were upregulated by 12.61, 11.60, 11.52, 10.78, 9.70, and 9.24 times in TR vs. MG, respectively (Table 2). Of these, ethyl-2-methylpropanoate and ethyl butanoate compounds are potentially the most important to mango aroma (Munafo et al., 2016). The KEGG pathway showed biosynthesis of secondary metabolites, sesquiterpene and triterpenoid biosynthesis, and metabolic pathways were the three most significant pathways (Figure 3C). Among the 19 volatile metabolites in Table 2, there were a total of 13 esters and terpenoids, of which six were upregulated and seven were downregulated in expression, and their expression abundance explained the slight difference in mango fruits aroma between MG and TR (Figure 3D).

The volatile organic compounds of plants are secondary metabolites playing an important role in fruit flavor. In mangoes, the volatile organic compounds mainly include esters, terpenes, alcohols, aldehydes, and acids. These compounds showed a declining trend in expression, indicating that the level of many volatile organic compounds declines sharply in MG fruit (Table 2). Among the 82 compounds detected in TR vs. MG (Supplementary Table S5), 50 were esters and 25 showed increases in TR relative to MG. Esters are synthesized mainly through the aliphatic acid pathway. In this study, eight important gene families were identified in the aliphatic acid pathway (Figure 4), namely, acetaldehyde dehydrogenase (ALDH, 36 members), cytochrome P450 (CYP, 33 members), flavin adenine dinucleotide (FAD, 23 members), hydroperoxide lyase (HPL, 2 members), alcohol dehydrogenase (ADH, 3 members), pyruvate decarboxylase (PDC, 18 members), Lipoxygenase (LOX, 6 members), and Arogenate dehydrogenase (AAT, 4 members; Supplementary Table S6). According to the expression analysis, it was found that the expression levels of FAD, HPL, and ADH decreased with fruit development, which is consistent with the decreasing trend of ester metabolite content. Among them, ADH is involved in the biosynthesis pathway of aroma volatiles in fruits by converting aldehydes to alcohols and providing substrates for ester formation (Manríquez et al., 2006). ALDH (30 and 6 genes showing significant up- and downregulation, respectively) and CYP (18 and 17 genes significantly up- or downregulated, respectively) indicating a general trend for gene upregulation than downregulation. CYP genes encode an enzyme that can catalyze a wide range of reactions but has been identified as responsible for the C10 hydroxylation of α-pinene to myrtenol, leading to different flavors in strawberry fruits (Aharoni et al., 2019). The differential expressions observed here in CYP family gene members may similarly impact on mango flavors in the transition from MG to the TR stage of fruit ripening. In addition, the other ester-related gene families of PDC, LOX, and AAT also showed differential regulations, suggesting that esters are important components of mango flavor substances.

Carotenoids are crucial players in photosynthesis and lipid peroxidation in plants and their content impacts on the color and appearance of fruits (Chisté et al., 2014). The dynamic changes in gene expression and differential metabolite levels in the carotenoid synthesis pathway in TR vs. MG were analyzed for a better understanding of the molecular genetics of mango pulp coloration (Figure 5; Table 3). The results of the analysis of significant DEGs indicated that 26 could be associated with changes in the carotenoid metabolome (Figure 5). Among them, the GGPS participating in carotenoid precursor synthesis showed a significant increasing trend in expression from MG to the TR stage. The 15-cis PSY genes playing an important role in the synthesis of β-carotene precursors all showed a similarly significantly increasing trend in expression. Among the PDS genes, three showed upregulated expression and one had a downregulated expression. The expression level of gene Cluster-16416.67635 encoding ZDS was significantly increased by 1.21 times in TR fruit. The expression level of LCYB gene, the key enzyme in carotene synthesis, was significantly upregulated by 3.26 times, indicating a rapid increase of β-carotene synthesis in TR. The results of carotenoid metabolome assays showed that the content of β-carotene was significantly increased by 3.31 times in TR vs. MG, suggesting that the high expression of LCYB gene can cause carotenoid accumulation. MG showed a faint yellow color which transitioned to orange in the ripening process and this can be attributed to the differences in the levels of α- and β-carotenes. CHYBs catalyze the synthesis of zeaxanthin from β-carotene or lutein from α-carotene and thus prevent the accumulation of carotenes. Among the CHYB genes, two were upregulated and one downregulated in MG fruit which might explain the enhanced accumulation of carotenes at the TR stage of ripening. Similarly, the upregulation in MG of all ZEP genes and three out of four NCED genes that operate down-stream of CHYBs in the β-carotene branch of carotenoid synthesis might also have contributed to relative decreased pool of β-carotene at the MG stage.

TFs are key players in controlling the expression of structural genes in secondary metabolite biosynthesis. In this study, 563 DEGs in TR vs. MG were found to be TFs. In “TR vs. MG,” the expression of 439 TFs was upregulated, while 124 TFs showed downregulation. TFs were annotated as NAC, WRKY, bHLH, MYB, AP2/ERF, bZIP, HSF, COP1, or MADS-box (Supplementary Table S7). The bZIP TF, HY5 (16416.43081), was downregulated by 1.82 times in TR vs. MG. In addition, 76 of important MYBs were selected from “TR vs. MG.” among which 56 genes were upregulated and 20 genes were downregulated in expression. 16416.57250, 16416.71538, and 16416.47912 were reduced by 8.90, 8.61, and 8.42 times, respectively (Supplementary Table S7). As vital regulatory TFs in carotenoid-biosynthesis pathways, bHLHs are involved in the regulation of important carotenoid-biosynthesis genes, such as LCYB. In “TR vs. MG,” 96 differentially expressed bHLHs were detected, 89 of them were upregulated, and 7 were downregulated in expression. NACs, the largest TF family in plants, are expressed in different tissues and at various stages of plant development. These participate in the regulation in plant growth and development and the plant responses to environmental stresses. In this study, 116 differentially expressed NACs were detected, with 73 showing upregulation and 43 with downregulated expression. 16416.58202 was annotated as NAC083. Its expression in TR was the highest, up to 165.27, showing a significant increase of 4.06 times compared with that in MG. Some studies have shown that CrMYB68 can directly interact with CrBCH2 and CrNCED5 leading to carotenoid accumulation (Zhu et al., 2020). CpbHLH1 and CpbHLH2 individually regulate the transcription of lycopene β-cyclase genes (CpCYC-B and CpLCY-B) during papaya fruit ripening (Zhou et al., 2019).

To validate the key results of the RNA-Seq, in this study, 13 carotenoid-biosynthesis pathway genes, 3 MYB, 2 bHLHs, and 2 NAC family TFs were selected, and their expression levels in MG and TR samples were analyzed using RT-qPCR (Supplementary Figure S3). The expression levels of these structural genes were in line with those of the RNA-Seq results.

The fruit flavor quality is mainly influenced by saccharides and organic acids involved in carbohydrate metabolism and the flavor of mangoes may be largely determined by the content and type of these different metabolites. The flavor of fruits is generally determined by the ratio and concentration of volatile compounds, such as phenols and alcohols, as well as sugars, organic acids, and amino acids. As for the changes in flavor that occur during fruit ripening, a few specific metabolites have been focused on including saccharides (fructose, glucose, and sucrose), organic acids, amino acids, and alcohols (Chen et al., 2009; Lombardo et al., 2011). Up to now, there are no studies on the effects of secondary metabolites on mango fruit flavor. In order to gain insight into metabolites contributing to mango flavor, broad-target metabolomics by HPLC-MS/MS was used to characterize the changes in metabolite levels that occur during the transition between the MG and TR stages of fruit development. A total of 603 related metabolites, such as sugars and organic acids, were identified, of which 309 metabolites were accumulated to a higher degree in MG mango fruit (Supplementary Table S8). A great deal of carbohydrates (including 20 saccharides) was identified and analyzed in mango pulps (Supplementary Table S4). Among these 20 saccharides, 14 saccharides exhibited increased concentrations at the TR stage. Among them, five increased saccharides (D-Glucoronic acid, D-Fructose-1,6-biphosphate, D-Ribose, Mannitol*, and Dulcitol*) were the main saccharides contributors of flavor in mango fruits. In addition, a total of 50 organic acids differentially expressed were identified, of which 34 organic acids showed significantly increased concentrations at the TR stage, partly contributing to the changes in fruit flavor. The constitution and richness of amino acids are key indexes of nutrition quality and are also important to the flavor (Choi et al., 2012). Among the 73 amino acids identified in this study, 45 displayed differential accumulations in TR vs. MG (Table 1) and four amino acids (Glutathione reduced form, O-phospho-L-serine, γ-glu-cys, and Hexanoyl-L-glycine) were increased in TR (Table 1), with a Log₂FC > 10. These results suggest that the differences in constitution and richness of amino acids also affect fruit flavor.

There were 143 volatile-related metabolites identified in this study, of which 82 volatile metabolites showed differential accumulation between MG and TR stages of mango pulp development (Supplementary Table S9). A large number of ester and terpenoid compounds were identified and their levels analyzed in this study. Pino and Mesa (2010) reported that ethyl-2-methylpropionate, ethyl butyrate, and methyl benzoate were the main esters determining the aroma of mango. Esters and terpenes are usually produced during fruit ripening by lipid metabolism; however, lipid metabolism is further enhanced by ripening temperature (Lalel et al., 2003). Esters are mainly synthesized by the fatty acid pathway, and eight important gene families were identified in this pathway (Figure 4). It was found that PDC, ALDH, FAD, and CYP gene families had large numbers of members: 18, 36, 23, and 35, respectively. These genes, especially ALDH genes, may play crucial roles in ester metabolism. Terpenoids have been reported to be the main fruit aroma substances, with terpinene having floral and sweet aroma properties, while γ-Terpinene and myrcene are responsible for the aroma of green mango fruits (Engel and Tressl, 1983; Tamura et al., 2001). Terpinene is the main volatile compound in cv. Willard and Parrot mango fruits (MacLeod and Pieris, 1984). In our study, γ-Terpinene was present as the main terpenoid in the “Golden Phoenix” mango fruit. Changes in the abundance of volatiles during ripening can determine the commercial value of mango fruits (Thiruchelvam et al., 2020). The type and concentration of volatile compounds vary considerably among cultivars. Interestingly, in our study, 22 terpenoids were upregulated in MG mango fruit. These differences explain the reason why MG fruits have a fruity aroma.

Mango fruit color is mainly composed from the flavonoids in vacuoles and the plastidial carotenoids and chlorophyll in pulps. In addition, another important factor for attractive mangoes is their fruit peel color which is determined by their contents of carotenoids, chlorophyll, and flavonoids (Karanjalker et al., 2018). There are more than 600 types of known carotenoids, which provide pigmentation to many plant organs, including fruits, and are vital players in consumer appreciation and economic values of plant products. In the ripening stage of apples, the background color is composed of chlorophyll, carotenoids, and flavonoids. The content and proportion of these three pigments determine the background color of the pericarps, thereby affecting the surface color (Mark et al., 2010). During mango ripening, its chlorophyll is almost completely degraded at 22°C. Anthocyanin contents also decrease slightly, while levels of carotenoids, mainly β-carotene and violaxanthin, are increased (Karanjalker et al., 2018). Furthermore, flavonoids are important pigments, which can produce a series of color from yellow to red and from faint yellow to orange (Ni et al., 2020). The transcriptomic analyses performed in this study identified some changes in the biosynthesis of flavonoid metabolite biosynthetic pathways between MG and TR fruit stages. From the metabolomics data, KEGG enrichment analysis also indicated that the concentration of metabolites in two phenolic pathways (“flavonoid biosynthesis” and “flavone/isoflavone biosynthesis”) in pulps was significantly different in fruits pulps at the MG and TR stages. It was confirmed through analysis of flavonoid metabolite profiles that the accumulation of flavonols, flavonoid, and isoflavone metabolites was increased in TR (Supplementary Table S4). In the biosynthesis pathway of phenylpropanoids, the expression levels of 4CL, C4H, CHI, and histohaematin P450 (FNS/IFS) genes presented positive correlations with the accumulation of flavonoids and isoflavones. Previous studies have shown that the structural genes in the phenylpropanoid pathway determine the accumulation of flavonoids and that these are partly regulated by MYB TFs (Supplementary Table S7). These, together with bHLH3 and WD40, can form MBW complexes, jointly leading to the biosynthesis of phenylpropanoids and proanthocyanin (PA; Nemie-Feyissa et al., 2015). In this study, many DEGs between MG and TR for MYB and bHLH TFs were detected, which may affect flavone/isoflavone biosynthesis during fruit ripening. However, their roles in this process require further experimental verification.