Anthocyanin and PA profiles of ‘CF’ and its red mutant ‘RCF’
Except for skin color, no significant morphological differences were observed between the fruits of ‘CF’ and its red mutant ‘RCF’. The young fruits of both ‘CF’ and ‘RCF’ initially had a deep green appearance. The color difference between ‘CF’ and ‘RCF’ became visible from about 5 days after full bloom (DAFB). The ‘RCF’ fruits quickly turned dark red, and retained their strong color until maturity. In contrast, the fruits of ‘CF’ only developed a slight red blush on the sun-exposed surface (Fig. 1A).
We identified and quantified the individual anthocyanins and PAs in ‘CF’ and its red mutant ‘RCF’. We identified and quantified 12 anthocyanins and seven PAs from ‘CF1’, ‘RCF1’, ‘CF2’, and ‘RCF2’. The anthocyanin metabolites included pelargonin, cyanidin 3-rutinoside, pelargonidin 3-Glu, malvidin-3-galactoside chloride, cyanidin 3-O-malonylhexoside, oenin, delphinidin, peonidin O-hexoside, cyanidin, and rosinidin O-hexoside; and the PAs included procyanidin A, procyanidin B3, procyanidin B, procyanidin A1, and procyanidin A2, which were all detected for the first time in pear. As shown in Fig. 1B, most anthocyanins and PAs were significantly up-regulated in ‘RCF’ compared with ‘CF’. The levels of anthocyanins and PAs in ‘CF’ and ‘RCF’ were initially low and then sharply increased during fruit coloration, except for delphinidin and rosinidin O-hexoside (Fig. 1B). These patterns of pigment accumulation were consistent with the strikingly different fruit color phenotypes of ‘RCF’ and ‘CF’.
The thresholds for significant differences in metabolite levels between the two cultivars were variable importance in projection (VIP) value ≥1 and︱log2(fold change)︱≥1. Against these criteria, six, eight, six, and eight anthocyanin metabolites, and four, seven, seven, and seven PA metabolites were significantly differentially accumulated in the four comparison groups: ‘RCF1’ vs. ‘CF1’, ‘RCF2’ vs. ‘CF2’, ‘RCF2’ vs. ‘RCF1’, and ‘CF2’ vs. ‘CF1’, respectively (Supplementary Table S1). Therefore, these anthocyanins and PAs were selected for further metabolite and transcript correlation analyses.
The RNA-seq process yielded 95.6 G clean bases and 637 million clean reads. The mean number of clean reads per sample was 53 million. Of the clean reads, 93.54% were mapped in total, and 90.72% were mapped uniquely against the improved apple reference genome sequence. In total, 14514 genes were expressed with FPKM ≥10 (Supplementary Table S2).
We identified 4065 DEGs in the four comparison groups. There were 340 DEGs in ‘RCF1’ vs. ‘CF1’, 252 in ‘RCF2’ vs. ‘CF2’, 2379 in ‘RCF2’ vs. ‘RCF1’, and 3055 in ‘CF2’ vs. ‘CF1’ (Supplementary Table S3): in those comparison groups, 123, 155, 858, and 1060 genes were up-regulated, and 217, 97, 1521, and 1995 genes were down-regulated, respectively (Fig. 2A). Only 12 DEGs were common to all four comparison groups (Fig. 2B).
The DEGs between group 2-1 and group RCF-CF were subjected to GO (Supplementary Table S4) and KEGG functional pathway analyses (Supplementary Table S5). For correlation tests with anthocyanins and PAs, we chose DEGs in group 2-1 categorized into DNA binding, plant hormone signal transduction, flavonoid biosynthesis, phenylpropanoid biosynthesis, flavonoid metabolism, phenylalanine metabolism, glutathione metabolism, and DEGs in group RCF-CF categorized into DNA binding, plant hormone signal transduction, phenylpropanoid biosynthesis, flavonoid biosynthesis, glutathione metabolism, and phenylalanine metabolism (Fig. 2C, D). In total, we selected 203 DEGs. Of these, there were 12 DEGs that overlapped between group 2-1 (‘RCF2’ vs. ‘RCF1’, and ‘CF2’ vs. ‘CF1’) and group RCF-CF (‘RCF1’ vs. ‘CF1’, ‘RCF2’ vs. ‘CF2’). These 12 DEGs encoded a β-glucosidase (PCP011059), two peroxidases (PCP024451 and PCP017906), a CHS (PCP023048), a homeodomain protein (PCP024513), an ERF (PCP044584), a RBR (retinoblastoma-related protein, PCP007207), an AUX (PCP036703), a NAC (PCP028501), a GST (PCP025171), and two SAURs (PCP037299 and PCP040169) (Fig. 2E). Of these, PcGST (PCP025171) was the most up-regulated gene in the comparison groups ‘RCF1 vs. CF1’ and ‘RCF2 vs. CF2’ (Fig. 2F).
Correlation analysis between selected transcripts and anthocyanins/PAs
To identify the candidate genes in anthocyanin and PA accumulation in pear, we conducted correlation analyses between selected transcripts and metabolites. In total, we detected 420 significant correlations (correlation coefficient, R2>0.8) between 203 transcripts and 17 metabolites, including 10 anthocyanins (kuromanin chloride, pelargonin, cyanidin 3-rutinoside, pelargonidin 3-Glu, malvidin-3-galactoside chloride, cyanidin 3-O-malonylhexoside, oenin, delphinidin, peonidin O-hexoside, cyanidin, and rosinidin O-hexoside) and seven PAs (procyanidin A, procyanidin A1, procyanidin A2, procyanidin B, procyanidin B1, procyanidin B2, and procyanidin B3) (Supplementary Table S6). Each metabolite was correlated with many different transcripts. Malvidin-3-galactoside chloride, oenin, and delphinidin were correlated with the fewest transcripts: five, seven, and three transcripts, respectively. Kuromanin chloride and pelargonin were correlated with the highest numbers of transcripts: 184 and 56 transcripts, respectively. Interestingly, kuromanin chloride and pelargonin shared the largest number of common transcripts (36 transcripts). This suggested that kuromanin chloride and pelargonin might have evolved similar accumulation mechanisms.
The 203 transcripts were annotated with descriptions from the SwissProt and NR databases. Six transcripts have been functionally characterized to play roles in anthocyanin accumulation in pear previously: PcMYB10, PcMYB114, PcCHS, PcCHI, PcF3H, and PcANS (Supplementary Table S7). The rest were newly identified as candidate genes involved in anthocyanin and PA accumulation in pear. The 203 transcripts were grouped into two clusters (I-II) (Supplementary Table S7). Genes in cluster I were strongly correlated with anthocyanins. Cluster I comprised 183 genes (90.1%). Of these, 147 genes were correlated with a single anthocyanin: 142 genes were correlated with kuromanin chloride, three genes were correlated with cyanidin, one gene was correlated with malvidin-3-galactoside chloride, and one gene was correlated with pelargonin. The remaining genes in cluster I were closely correlated with two or more anthocyanins: 31 genes were commonly correlated with kuromanin chloride and pelargonin, two genes were correlated with pelargonin and cyanidin, two genes were correlated with malvidin-3-galactoside chloride and oenin, and one gene was correlated with cyanidin 3-rutinoside, oenin, and cyanidin. Cluster II contained 20 genes (9.9%) that were strongly correlated with both anthocyanins and PAs. Of these genes, two phenylpropanoid structural genes (encoding 4CL1 and 4CL2), six flavonoid structural genes (encoding CHS, 3 CHIs, F3H, and ANS), six TF genes (encoding bZIP1, MYB3, MYB86, MYB111, MYB114, and KNAT1), two phytohormone signal transduction genes (encoding IAA13 and ERF003), two DNA-directed RNA polymerase genes (encoding rpoB and Rpb1) and one GST transporter gene (encoding GSTF12) were positively correlated with anthocyanins and PAs. One WRKY TF gene, WRKY28, was negatively correlated with anthocyanins and PAs (Supplementary Table S7, S8). Each gene in cluster II was strongly correlated with many metabolites. We found that these 20 genes were strongly correlated with all 17 anthocyanin and PA metabolites (Fig. 3A). Thus, they were considered to represent the core genes for anthocyanin and PA accumulation in pear. Of these, PcRPB1 (PCP004386) was correlated with the fewest metabolites: one anthocyanin and three PAs; and PcGSTF12 (PCP025171) was correlated with the most metabolites: seven anthocyanins and seven PAs (Fig. 3B).
qPCR analysis of DEGs related to anthocyanin and PA accumulation
To validate the RNA-seq data, we conducted qPCR analyses of 10 of the anthocyanin and/or PA candidate genes: PcCHI, PcC1, PcMYB114, PcHB7, PcGAI1, PcCHS, PcGSTF12, PcANS, PcHB12, and PcMYB10 (for gene IDs and primers, see Supplementary Table S9). The transcript profiles of all selected genes were very similar to those detected from the RNA-seq data (Fig. 4). The results showed that PcGSTF12 was most up-regulated in comparison groups ‘RCF1 vs. CF1’ and ‘RCF2 vs. CF2’. This result was highly consistent with the results of RNA-seq, and provided further evidence for the crucial role of PcGSTF12 in anthocyanin and PA accumulation in pear. Thus, we conducted further analyses to confirm the function of PcGSTF12.
PcGSTF12-mediated anthocyanin and PA accumulation in pear
Our combined metabolite and transcriptomic analyses revealed a core set of genes closely correlated with pear anthocyanins and PAs, which strongly suggested that they play key roles in anthocyanin and PA accumulation in pear. To test this, we focused on the most up-regulated gene among the core set of anthocyanin and PA candidate genes, PcGSTF12, for functional analysis.
(1) Functional analysis of PcGSTF12
The phylogenetic analysis showed that PcGSTF12 is a homolog of FvRAP in strawberry, Riant2 in peach, and MdGST in apple, all of which are in the phi subfamily  (Fig. 5A). Members of the phi subfamily are anthocyanin transporters. To test the potential role of PcGSTF12 in anthocyanin accumulation, 35S:: PcGSTF12 was transformed into the Arabidopsis mutant tt19-7 (for primers, see Supplementary Table S9). The tt19-7 plants showed a green hypocotyl phenotype, while the tt19-7-OE transgenic plants showed the red hypocotyl phenotype, like that of the wild type (WT) (Fig. 5B). However, the brown color of seed coats was not rescued in the tt19-7-OE lines (Fig. 5B). This result was consistent with the fresh seed phenotype obtained by transferring 35S:: RAP-RFP into Arabidopsis tt19-7 .
To explore the role of PcGSTF12 in anthocyanin and PA accumulation, we conducted a metabolite analysis using Arabidopsis seedlings. Three PAs and nine anthocyanins were significantly up-regulated, and three anthocyanins were significantly down-regulated in tt19-7-OE compared with tt19-7. Procyanidin A3, cyanidin O-acetylhexoside, delphinidin 3-O-rutinoside, and cyanidin 3-p-hydroxybenzoylsophoroside-5-glucoside were specifically up-regulated in tt19-7-OE compared with tt19-7 (Fig. 5C). Five anthocyanins (malvidin 3-acetyl-5-diglucoside, pelargonidin 3-O-beta-D-glucoside, delphinidin 3-O-rutinoside, pelargonin, cyanidin 3-O-galactoside) and one PA (procyanidin A3) were significantly up-regulated, and two anthocyanins (petunidin 3, 5-diglucoside and delphinidin O-malonyl-malonylhexoside) were significantly down-regulated in tt19-7-OE compared with WT. Interestingly, a large amount of petunidin 3, 5-diglucoside was detected in WT but not in tt19-7-OE. In contrast, procyanidin A3 was detected only in tt19-7-OE. These results confirmed that PcGSTF12 is responsible for anthocyanin and PA accumulation. Interestingly, its affinity for anthocyanins and PAs differed from that of AtGSTs in Arabidopsis. In particular, our results showed that PcGSTF12 is responsible for the accumulation of procyanidin A3 but not petunidin 3, 5-diglucoside, opposite to the function of AtGSTs in Arabidopsis. PcGSTF12 is a newly identified member of the phi GST family involved in anthocyanin and PA accumulation.
Next, we analyzed RNA-seq data to identify which genes were affected by PcGSTF12 in the seedlings of tt19-7-OE vs. tt19-7. In total, we found 28 strongly affected genes, which encoded proteins involved in anthocyanin and PA biosynthesis, regulation, and transport (Supplementary Table S10). These results showed that PcGSTF12 might not only be an anthocyanin and PA transporter, but may also participate in many other steps of anthocyanin and PA accumulation.
(2) Upstream regulation of PcGST12
Correlation analyses showed that the transcript level of PcGSTF12 was significantly correlated with that of PcMYB114 (Fig. 5D). Further, several MYB-binding sites were found in the PcGSTF12 promoter, indicating that PcGSTF12 might be directly bound by, and regulated by, MYB transcription factors (Supplementary Table S11). Several R2R3-MYB genes are known to bind to MBS sites , and we found a MBS site within an 801-bp region upstream of the start codon. Thus, this MBS site was used in an EMSA assay (for primers, see Supplementary Table S9). The biotinylated probe was able to bind PcMYB114 protein, and the addition of a high concentration of cold probe significantly reduced the binding affinity of the biotinylated probe. To test whether PcGSTF12 could be regulated by PcMYB114, we further carried out luciferase reporter assay (for primers, see Supplementary Table S9). The relative LUC activity of the PcGSTF12 promoter was about 6-fold that of the empty vector control. These results showed that PcMYB114 could directly bind to the MBS site in the PcGSTF12 promoter (Fig. 5E) and positively regulate its activity (Fig. 5F).
We also found many cis-acting elements involved in auxin-, ethylene-, and gibberellin-signaling in the PcGSTF12 promoter. This indicated that PcGSTF12 might be the common downstream target of R2R3-MYBs, and auxin, ethylene, and gibberellin signals that regulate the anthocyanin and PA pathways (Supplementary Table S11). Together, these results provide new clues about PcGSTF12-mediated anthocyanin and PA accumulation in pear.