**3. Discussion**

Anthocyanins and flavonoids are the key pigments that determine flower colors. In general, cyanidin, pelargonidin, and paeoniflorin are beneficial to the formation of red color; delphinium and malvidin are conducive to the formation of blue color [9,33]. The combined effect of these pigments makes the flowers of *Rhododendron* present white, pink, red, purple, and other colorful colors. The results of HPLC show that the main anthocyanins in different samples were the same, but the content of each pigment was significantly different. It is inferred that the color depth of *R. pulchrum* is related to the content of main anthocyanins in the petals. The high content of peonidin in sample A may be an important factor in the formation of deep color. Peng et al. also put forward a similar conclusion that the anthocyanin biosynthesis pathway is the main metabolic pathway of flower color formation in their study on *Hydrangea macrophylla* cv. 'Forever Summer' [34]. The biosynthesis and floral color regulation mechanism of flavonoids in *R. pulchrum* was elaborated in Xia et al.'s research [12]. In the four samples of this experiment, flavonoid content was negatively correlated with colors. This shows that the color depth of *R. pulchrum* is produced by the combined action of anthocyanins and flavonoids, which is consistent with the research results of Liu et al. [35].

The intensity of petal color change was correlated with the concentration of anthocyanin in plants [36]. Xue [37] reported that the anthocyanin content increased as the flower color deepened, and the expression of key structural genes was significantly increased in red-flowered strawberry in comparison with white-flowered strawberry. Previous results suggested that the competition expression of FLS and DFR genes led to the redirection of flavonol and anthocyanin accumulation [38,39]. In this study, the difference in gene expression was consistent with the accumulation of anthocyanins in different samples. This discovery is like that of blueberry fruit development, in which a partial correlation between anthocyanin-related gene expression and anthocyanin concentration was found [40]. Meanwhile, transcriptome analysis revealed that the FLS and DFR genes were differentially expressed in the flower color of the four different varieties. Six genes encoding FLS were highly expressed, especially in the D sample, and three genes in the A sample associated with DFR were more highly expressed than in the B, C, or D samples. These results indicate that the expression levels of genes encoding DFR were higher in deeper color flowers, which significantly promoted the conversion of dihydroflavonols to anthocyanins. On the contrary, the increased expression of FLS in the white flowers promoted the conversion of dihydroflavonols into flavonols (Table 4). There may be substrate competition between FLS and DFR, controlling the metabolic flux via the branching of the flavonoid biosynthetic pathway, resulting in changes in flower color intensity in *R. pulchrum*.

Transcription factors perform important roles in all plants. In the current study, most of the TFs were enriched in ERF, MYB, *bHLH*, WRKY, MYB related, FAR1, and NAC families during anthocyanin transformation (Figure 7b). ERF plays a particularly important role in the color expression of light petals, especially pink petals, by affecting the upstream pathway of anthocyanin biosynthesis [41]. The class of TFs was previously involved in the regulation of petal color formation [42]. The *bHLH* proteins, forming one of the largest TF families, play vital roles in various metabolic, physiological, and developmental processes in plants [43]. In our study, one gene (*Rhsim08G0230000*) encoding *bHLH* was more highly expressed in the A sample, indicating that this *bHLH* was a positive regulator of anthocyanins biosynthesis in *R. pulchrum.* It has been reported that MYB is closely related to flavonoid metabolic pathways [44]. However, some studies showed that there were also inhibitory factors in the MYB family that inhibit anthocyanin biosynthesis [45]. In our study, one gene (*Rhsim09G0042000*) encoding the MYB family members was significantly down-regulated in the D vs. A, C vs. A, and B vs. A comparisons (Table S5), which suggested it inhibited the anthocyanin accumulation.

Phytohormones are also critical internal factors affecting anthocyanin biosynthesis in many plant species. Wang et al. [46] reported that auxin inhibits anthocyanin accumulation and reduces the expression of genes linked to anthocyanin biosynthesis in apples. Based on our transcriptome data (Table S6), six genes associated with auxin-mediated signaling pathways were detected in the three groups, four of which were down-regulated in the A sample. In addition, the transcript level of two genes (*Rhsim09G0181000* and *Rhsim07G0028200*) encoding brassinosteroid was increased in the A sample. Similarly, Peng et al. [47] found that BR affects anthocyanin accumulation by regulating the anthocyanin biosynthesis genes in *Arabidopsis* seedlings. In addition, the response proteins in signal transduction of IAA and ABA were prominently regulated during anthocyanin transformation, and the functions of these hormone-responsive proteins deserve further investigation.

The results for anthocyanin composition in petals and deep-color blotches obtained by HPLC show that the accumulation of delphinidin and peonidin may be the pigment basis for the formation of deep-color blotches (Table S1). In the four samples, the downregulation of DFR genes related to anthocyanin synthesis (Table 4) was inconsistent with the decreasing trend in anthocyanin contents. This is possibly due to the defect that petals and deep-color blotches were not sequenced separately for RNA-seq. The presence of deepcolor blotches resulted in high expression of DFR genes in B and C samples. Studies of tree peony (*Paeonia suffruticosa*) have already shown that petal coloration and spot coloration are differentially regulated by MYB transcription factor [25,48]. Separating petals and deep-color blotches for independent analysis, revealing the transcriptional and metabolic regulatory mechanisms of petal and deep-color blotches formation, will be the focus of our subsequent research.
