The Transcriptome of Dahlia pinnata Provides Comprehensive Insight into the Formation Mechanism of Polychromatic Petals

Abstract

Garden dahlias (Dahlia pinnata) are popular for their wide range of color variations, with polychromatic cultivars enhancing their ornamental value. Previous studies on the anthocyanin biosynthetic pathway (ABP) have indicated that the post-transcriptional suppression of the chalcone synthase gene (CHS) is involved in the formation of the white petals of dahlias. To further explore the complex mechanisms underlying polychromatic petal formation, we selected the bicolor cultivar ‘LiRen’ to identify candidate genetic factors. Through the detection of proanthocyanidin and anthocyanin, it was indicated that the white tips of the petals lacked anthocyanin but accumulated some proanthocyanidin, albeit at significantly lower levels than those at the red bases of the petals. This suggests that the upstream ABP, which involves CHS, is not entirely inactive. Transcription sequencing and quantitative reverse transcription PCR (qRT-PCR) analysis demonstrated that the inactive ABP in the white tips results from the downregulation of ABP structural genes. The low abundance of DpMYB1 appears to be the key factor influencing the lack of strong transcription activation of the structural genes. Additionally, highly upregulated DpSPL9 targeted by the downregulated miR156 in the white tips was identified through qRT-PCR. This suggests that DpSPL9 may act as an anthocyanin depressor to destabilize the MYB-bHLH-WDR complex through interaction with DpMYB1. The findings indicate that the DpMYB1 and miR156-DpSPL9 modules play potential regulatory roles in the formation of bicolor petals. Overall, these results provide new insights into the color patterning of dahlias and will be valuable for further studies regarding the mechanisms underlying polychromatic petal formation.

1. Introduction

Dahlias, perennial herbaceous plants of the Asteraceae family, are widely cultivated as ornamental plants in many countries due to their diverse inflorescence shapes, colors, and sizes, as well as their long flowering period [1,2,3]. The dahlia cultigen (garden dahlias), i.e., Dahlias pinnata (syn. variabilis) [4], has produced over 57,000 cultivars, including many polychromatic varieties that are autoallooctoploids (2n = 8X = 64) resulting from a doubled allotetraploid [3]. The rich array of flower colors, including ivory [2], yellow [1], red, purple [5], and black [6], contrast with a wide range of white tips in ray florets, creating bicolor or star-type dahlias with significant ornamental value. The bases of ray florets accumulate rich flavone pigments, primarily cyanidin and pelargonidin derivatives, which are lacking in the white tips [2,7]. In addition, flavones and their derivatives (apigenin and luteolin) are accumulated in white-colored commercial cultivars whose ivory petals are different from the pure-white color of the petal tips of bicolor or star-type dahlias [2,8,9].

Previous studies have indicated that the petal color arises from the accumulation of flavonoids, particularly anthocyanins, which are glycoside compounds of anthocyanidins (e.g., cyanidin, delphinidin, malvidin, peonidin, pelargonidin, and petunidin). Flavonoid biosynthesis begins with the participation of three key enzymes in the general phenylpropanoid pathway: phenylalanine ammonia-lyase, cinnamate 4-hydroxylase, and 4-coumarate-CoA ligase (PAL-C4H-4CL). PAL-C4H-4CL converts phenylalanine into p-couumaroyl-coenzyme A (p-couumaroyl-CoA). These enzymes convert phenylalanine into p-coumaroyl-CoA, which serves as the precursor for flavonoid synthesis. In the flavonoid biosynthesis pathway, p-couumaroyl-CoA is converted to naringenin chalcone (the yellow color) by chalcone synthase (CHS), followed by isomerization to naringenin (colorless) via chalcone isomerase (CHI). Naringenin is subsequently catalyzed into flavanonol (e.g., dihydrokaempferol, dihydromyricetin) by flavanone hydroxylase, which includes flavonoid-3′,5′-hydroxylase (F3′H) and flavanone 3-hydroxylase (F3H), and is then reduced to leucoanthocyanidin (e.g., leucodelphinidin and leucocyanidin) by dihydroflavonol reductase (DFR). Finally, leucoanthocyanidin is transformed into colored anthocyanidins through the action of anthocyanin synthase (ANS), while the most unstable anthocyanidins are converted into stable anthocyanins by various modification enzymes, including glycosyltransferases (e.g., anthocyanidin 3-O-glucosyltransferase, and GT/BZ1), methyltransferases (MT), and acyltransferases (e.g., anthocyanidin 3-O-glucoside 6″-O-acyltransferase, 3AT). Collectively, the upstream PAL-C4H-4CL pathway, the flavonoid biosynthesis pathway, and the modification of enzymes regulate the process of anthocyanin biosynthesis, referred to as the anthocyanin biosynthetic pathway (ABP) [7,9,10]. Besides the ABP, other flavonoids or derivatives of the flavonoid biosynthesis pathway share common and early biosynthetic genes (EBGs), including CHS, CHI, F3H, and F3′H [11]. Flavones (e.g., apigenin and luteolin) are synthesized by flavone synthase (FNS) [6], which stops the anthocyanidin metabolic flux induced by the laterbiosynthetic genes (LBGs) DFR and ANS [12]. In the downstream pathway, leucoanthocyanidin reductase (LAR) and anthocyanidin reductase (ANR) produce proanthocyanidins (PAs) using the same precursor as that used for anthocyanins.

In addition to the structure genes, transcription factors (TFs) such as R2R3-MYB and bHLH form a highly conserved protein complex with the WD40 repeat protein, also known as ’MBW’ [13,14], to activate multiple ABP genes across all angiosperms. Notable examples include the ‘PAPs- or TT2-TT8-TTG1’ complex of Arabidopsis and the ‘AN2-AN1-AN11’ complex of Petunia hybrida [12,13]. In dahlias, DvIVS (TT8/AN1 homolog) plays a crucial role in regulating the ABP through activating DvCHS1, DvF3H, DvDFR, and DvANS. However, transposable element insertion leads to the loss of the bHLH domain in DvIVS, resulting in a lack of anthocyanin in bright yellow ray florets [7,15]. Though the expression of DvIVS and DvCHS is essential for anthocyanin accumulation, polychromatic petal formation depends on the special pigmentation pattern, which is impacted by MBWs and other regulators [13,14]. Anthocyanin repressors serve as fine regulators that control pigmentation intensity and location, either by actively repressing MBW or passively destabilizing MBW. Active repressors, such as SG4 R2R3-MYBs, suppress the promoter activities of the ABP or the phenylpropanoid pathway (PAL-C4H-4CL) [16]. Many SG4-R2R3 MYBs induced by MBW positively correlate with the anthocyanin content and seem to provide feedback inhibition, which limits the anthocyanin biosynthesis intensity [7,15,16]. Conversely, positive repressors like CPC, an R3-MYB activated by MBW, enhance the formation of inactive MBW in adjacent cells through intercellular movement. This fine-tuning of pigmentation by CPC-like R3-MYBs leads to the formation of anthocyanin spots and color intensity in petals [13,14,17]. Besides CPC [18], other TFs, such as SPL [19], NAC [20], LBD [21], and HD-ZIP [20], negatively regulate MBW via protein–protein interactions (PPIs) with MBW components. Competitive PPI between PAP1 and SPL9 destabilizes the MBW complex and inhibits the ABP. Moreover, the age-dependent pathway, which involves the miR156-SPL model, allows SPL9 to integrate anthocyanin production with vegetative-phase transitions [16].

Previous studies have suggested that the siRNA-mediated post-transcriptional gene silencing (PTGS) of CHS or a lack of full-length IVS may contribute to anthocyanin absence in the ray florets of polychromatic dahlias [2,7]. MBW and various repressors are responsible for different pigment patterns of flowers through complex regulatory mechanisms [16]. This indicates that regulatory TFs play a key role in fine-tuning bicolor petal formation, a simple form of differential pigmentation. To investigate the key TFs and their potential roles in transcriptional regulation, ray florets of a bicolor D. pinnata cultivar were selected for gene expression comparison. Transcriptome sequencing (RNA-seq), along with assembled unigenes and their expression profiles, provided cues for the prediction of candidate TFs responsible for the white tips and red bases of ray florets. Based on the ABP and MBW frameworks, candidate TFs were identified to construct a transcription regulation map for dahlias.

2. Materials and Methods

2.1. Plant Materials, PA Content Detection, and Unigene Assembly

Flowers of bicolor D. pinnata ‘LiRen’ (commercial cultivar, Figure 1a) were collected from the cutting seedings grown in greenhouse of YanXi Flowers Co., LTD (Yangzhou, Jiangsu, China). The red bases of ray florets (R) and white tips of ray florets (W) (Figure 1b) were collected, respectively, and immediately frozen in liquid nitrogen.

PA of W and R were quantified using Plant Oligomeric Proantho Cyanidins Kit (A144-1-1, Jiancheng, Nanjing, China). According to the manufacturer’s instruction, the standard curve based on the PA concentration gradient and absorbance of 500 nm (OD₅₀₀) was created, then PA of W and R were extracted, and PA contents were calculated by corresponding OD₅₀₀.

Total RNAs of W and R (2 repetitions each) were extracted using RNAprep Pure Plant Kit (DP441, Tiangen, Beijing, China). Transcriptome library construction and sequencing were performed on NovaSeq 6000 platform (Illumina, San Diego, CA, USA) through an outsourcing service of the Biomarker Biotechnology Co., LTD (Beijing, China). Clean reads were collected from the RNA-seq raw data. Transcripts were obtained via de novo assembly with the 25 bp k-mer dictionary of Trinity [22]. Following redundancy elimination, unigenes and deductive protein sequences were obtained for further analysis.

2.2. Gene Annotation, Expression Analysis, and Differential Expression Analysis

All unigenes were BLASTed in Nr, Swiss-Prot, PFAM, KOG, GO, COG, KEGG, and eggNOG databases to obtain gene annotations. Using all unigenes as the reference sequence, the clean reads were mapped to reference to obtain the read count of each unigene via Hisat2 software (https://daehwankimlab.github.io/hisat2/ (accessed on 1 February 2024)). Gene expression levels were estimated by fragments per kilobase per million mapped reads (FPKM) according to the read counts of unigenes using featureCounts program in Subread software (https://subread.sourceforge.net/ (accessed on 1 February 2024)). Differentially expressed unigene (DEU) identification was performed using DEGseq2 software (https://github.com/thelovelab/DESeq2 (accessed on 20 February 2024)) with the threshold of ‘q-value < 0.01’ & ‘|log2(fold change)| > 1’ while referring to other omics studies.

2.3. Identification of ABP Structure Genes and Regulation Genes

The unigenes of ABP were identified via KEGG pathway mapping and gene annotation of KEGG. According to the known genes of MBW and the ABP depressors in Arabidopsis, Petunia hybrida, Zea mays, and D. pinnata ‘Yuino’, homologous unigenes were identified via Blastn and BlastP with E-value cutoff of 1e-5.

2.4. Identification, Phylogenetic Analysis, and Motif Analysis of DpSPLs

The DpSPL gene candidates were predicted from the deductive proteins using hidden Markov model (HMM) of SBP domain (pfam03110). Any candidate genes without SBP-boxes were rejected using TF prediction tool of PlantTFDB. Additionally, reference sequences of AtSPL family of Arabidopsis were obtained from PlantTFDB (Table S3). Multiple sequence alignment of the SBP-boxes of DpSPLs and AtSPLs based on MUSCLE algorithm and neighbor-joining phylogenetic analysis with 1000 bootstrap replicates were performed using MEGA 7 [23]. Top 10 conserved motifs were identified using MEME (http://meme-suite.org/tools/meme (accessed on 2 May 2024)) with default parameters. The targeted sites of miR156 within CDS of SPL genes were predicted using psRNATarget (https://www.zhaolab.org/psRNATarget/ (accessed on 10 May 2024)).

2.5. Validation of Expression Level of Key Genes Using qRT-PCR

Each 1 μg total RNA of W and R was reverse-transcribed using the PrimeScript RT reagent kit (RR036Q, Takara, Dalian, China). qRT-PCR was performed on Viia 7 Real Time PCR System (ABI, Foster, CA, USA) with EvaGreen Mix (Biotium, Fremont, CA, USA), following recommended reaction mixture (20 μL) and amplification procedure. The gene-specific primers and the stem-loop primer of miR156 are listed in Table S6.

3. Results

3.1. The Unigene Assembly, Function Annotation, and Expression Analysis

A total of 172,202 transcripts with an N50 length of 1485 bp were assembled from 24.28 GB clean reads generated from 4 transcriptome libraries (6.4, 6.2, 5.8, 5.7 GB for W1, W2, R1, R2, respectively). After filtering redundant sequences, 77,244 unigenes with an N50 length of 1308 bp were selected for function annotation and read count analysis. The multiple sequence databases (NR\Swiss-Prot\COG\KOG\eggNOG\GO\KEGG\PFAM) annotated 42,533 unigenes, with 5729 co-annotated across all 8 databases (Figure 2A). The density distribution of logarithmic FPKM values obeyed a normal distribution ranging from a distribution interval of −1 to 3 (FPKM 0.1–1000) with a mean value of 1 (FPKM = 10). The parallel quartiles and medians of FPKM values indicated that the global expression profiles of R and W were similar (Figure 2B,E). Pearson’s correlation coefficient of FPKM demonstrated higher similarity among biological replicates within each group compared to the similarity between groups (Figure 2C). This suggests that the biological replicates of R and W are reliable for reducing biological variability in differential expression analysis. A total of 792 DEUs were identified, and 621 of them were downregulated in the W vs. R comparison (Figure 2D, Table S1). GO enrichment analysis indicated that transferases were highly active, with several enriched GO terms, including acyltransferase activity, transferring groups other than amino-acyl groups (GO: 0016747), and UDP-glycosyltransferase activity (GO: 0008194), which could be involved in anthocyanin modification (Figure 2F).

3.2. The Depression of ABP Pathway

KEGG enrichment analysis of DEUs revealed that the top 10 enriched pathways included the phenylpropanoid biosynthesis pathway (ko00940) and anthocyanin biosynthesis pathway (ko00941), both of which were significantly less active in W compared to R (Figure 3A). According to the gene annotations of ko00940 and ko00941 (Table S2), 21 DEUs coding 10 key enzymes of the ABP were identified as crucial structure genes. For each enzyme, the most significant DEU (Table S2, the bolder unigenes) was selected for qRT-PCR validation. The relative expression levels of these unigenes closely correspond to fold-change values (Figure 3B). Pathway mapping illustrates the downregulation of corresponding genes across three steps of the general phenylpropanoid biosynthesis pathway and seven steps of the ABP. Significant repression of gene expression in W was observed in EBGs and LBGs, with the exception of F3′H and the PAL-C4H-4CL module (Figure 3C). This suggests that the total metabolic substrate availability for ABP in W was reduced compared to R, potentially inhibiting anthocyanidin accumulation in the white tips. Notably, the 18.4-fold downregulation of CHS, which represents the initial step of the ABP, and the 39.4-fold downregulation of 3AT, responsible for the terminal steps of ABP, explain the observed anthocyanidin deficiency in W.

No anthocyanidin was detected in W, consistent with findings from previous studies [2,7]. While a low abundance of CHS and 3AT significantly reduced anthocyanidin biosynthesis, alternative pathways, such as those for flavones, flavonols, and PAs, which utilize metabolic intermediates from the ABP, could also contribute to the competitive inhibition of anthocyanidin production [2,7]. The PA content in W (12.29 mg/g) was 15.2% of that in R (81.05 mg/g), while flavones or flavonols were not detected in W (Figure 3C). Interestingly, flavone synthase (FNS) and flavonol synthase (FLS), which are responsible for flavones and flavonol production, did not exhibit downregulation. However, the abundance of ANR-producing PAs was significantly lower in W (Figure 3C).

3.3. Identification of MBWs and Repressors of ABP

The downregulation of most ABP structure genes indicated board transcriptional repression in W. A homology search of well-known MBW genes (

Table 1. Candidate dahlia gene homologs to plant anthocyanin biosynthesis pathway regulators.

Species	R2R3 MYB		bHLH		WD40
	SG6	SG5	R	IN	TTG1-Clade	MP1-Clade
Arabidopsis thaliana #	PAPs	TT2 B	GL3 EGL3	TT8	TTG1	AN11A
Petunia hybrida #	AN2	\	JAF13	AN1	AN11	\
Zea mays #	\	C1	R	IN1	IN1	MP1
Dahlia pinnata (GeneBank Ac.)	MYB1 (AB601003)	TT2-like	DEL (AB601006)	IVS (BAJ33520)	WDR1 (AB601007)	WDR2 (AB601008)
Homologous unigenes	DN189_c1_g1	DN185_c0_g2	DN7019_c1_g1	DN402_c0_g1	DN14226_c0_g1	DN7103_c0_g2
	DN434_c0_g1	DN22158_c0_g2

# Genebank accession numbers of corresponding species listed in [16,17].

) identified candidate genes of dahlias, including two MYB genes (MYB1, a PAPs homolog, and TT2-like, an SG5 R2R3-MYB), two bHLH genes (DEL, a GL3 homolog, and IVS, a TT8 homolog), and two WD40 genes (WDR1, a TTG1 homolog, and WDR2, an AN11 homolog). Among these, only MYB1 significantly downregulated from a higher abundance in R (FPKM = 449.7) to a lower abundance in W (FPKM = 100.5), while the expression levels of other candidate genes remained low or similar between the two petal regions (Table S1). In addition to the reduced abundance of MYB1, ABP repression may be exacerbated by the MBW repressors through protein interaction, as observed with well-known CPC (R3 MYB). The equal abundance of CPC homolog (DN434_c0_g1) in R and W (FPKM = 132.3 and 142.1) suggests a stable background level of MBW repressors in the petals. Other MBW repressor candidates, such as LBD37/38/39 homologs (DN379_c2_g1/DN1309_c0_g1/DN379_c2_g2), the HY5 homolog (DN38191_c0_g1), and the GL2 homolog (DN9702_c0_g2) also exhibited uniform expression between R and W. Beyond passive repressors, active repressor such as SG4 R2R3-MYBs are known to directly repress ABP or general phenylpropanoid pathway genes via MBW-independent mechanisms [16]. The significantly lower abundance of the AtMYB6 homolog (DN8994_c0_g1) in W (FPKM = 3.7, while FPKM = 23.4 in R) seemed to result from weaker promoter activation by MBWs and does not contribute to the ABP depression observed in W (Table S1).

3.4. Identification of SPL Depressor Targeted by miR156

Another well-known MBW depressor, the SPL transcription factor, elucidates the molecular basis of juvenile reddening in seedlings via the miR156-SPL model [16]. A total of 15 DpSPLs were identified from the assembled unigenes (Table S3). SBP-box domains predicted using the NCBI Conserved Domain Search Service display conserved motifs linked by two zinc finger structures (Figure 4A, Zn-1/2). The CCCH motif of Zn-1 was fully conserved across all DpSPLs, while the CCHC motif exhibited high conservation, except for the His residue of DpSPL2/9 and the final cysteine residue of DpSPL9. Additionally, a nuclear localization signal motif located downstream of Zn-2, comprising four overlapping residues, was identified as a conserved feature, as it has also been identified in other plants [19].

A neighbor-joining tree of the SBP-box sequences revealed the conservation of DpSPLs and their corresponding homologous AtSPLs (Figure 4B). Based on the topological structure, all SPL genes were divided into seven groups (I-VII). Except VI, including the smallest SPL proteins (fewer than 181 AA), and I/VII, including the largest SPL proteins (more than 590 AA), the remaining four groups encoded SPL with lengths ranging from 200 to 430 AA. Furthermore, AA sequence alignment indicated that DpSPL16 (383 AA) has a significantly shorter C-terminal region compared to homologous SPL proteins in group VII, likely due to the translation of an incomplete coding sequence from the unigenes (Table S4). Notably, the conserved phylogenetic relationship suggests that DpSPL5 is absent from the unigene assembly, indicating that it may be a silenced gene in the petals of D. pinnata ‘LiRen’.

SPL silencing via miR156 overexpression in both Arabidopsis and poplar has led to ectopic anthocyanin accumulation in stems [16,17]. Seven DpSPL members containing miRNA response elements (MREs) are predicted as miR156 targets (Table S5). Motif analysis revealed that the composition of SPLs within the same group is highly conserved. MEME analysis (Figure 4B) indicated that MRE motif ‘ALSLLS’ (Motif 9) located within CDSs of all AtSPLs in III/IV/V is a conserved motif characteristic of miR156-targets branches. The MRE motif is located about 100 AA downstream of SBP-box (Motif 2-1-3) for DpSPL6 of III, DpSPL2/10/11 of IV, and 12/13 of V, except DpSPL1 (only 40 AA). It appears that an MRE should located on the 3′ terminal of DpSPL9, though only an incomplete CDS was available. Sequence alignments of SPLs in III confirmed the absence of the the MRE motif in DpSPL15 (Figure 4B, star), indicating that the CDS of DpSPL15 has lost the miR156 target site. It is notable that besides the conserved ‘ALSLLS’-type MRE of most SPLs, MRE isoforms ‘A[P/R]SLLS’ and ‘VLYLLS’ were identified in SPL1/6/12. ‘VLYLLS’ was coded by a target site of SPL12 whose ‘A’ site failed to complementally pair with the seed region of miR156 (the 11th nucleotide) (Figure 4C). This mismatch could result in translation inhibition, which was not observed in AtSPLs (Table S5). To validate the interaction, the abundance of full mRNA and cleavage fragments of mRNA were reverse-transcribed for abundance validation via qRT-PCR. The abundance of cleavage fragments of SPLs indicated that SPL2/6/9/10/13 targets via miRNAs in petals, and the interaction of miR156 with SPL9 was the most significant module (Figure 4D).

SPL family analysis provided seven candidates as MBW depressors. The gene expression profiles (Figure 5A) revealed significant fold changes in DpSPLs. The high abundances of SPL9/14 suggest their significant roles in pigmentation, while the low abundance of SPL4/12/13 excluded them as MBW depressor candidates. The upregulation of SPL9/12 in W, contrasted with the downregulation of SPL1/3/8/15, provides further insights for narrowing potential depressors. DpSPL9 exhibited an 8.5-fold increase in W, transitioning from low abundance in R to high abundance in W, supporting its role as a strong MBW depressor candidate. Furthermore, significant cleavage of DpSPL9 by miR156 in R (Figure 4D) indicated a correlation between DpSPL9 and miR156 abundance in petals. The reduced abundance of miR156 in W suggests that the desilencing of DpSPL9, contributed to its increased expression (Figure 5B). Collectively, these findings indicate that miR156-induced desilencing of DpSPL9 may depress anthocyanin accumulation in the white tips by competitively interacting with PAP1, a component of MBW.

4. Discussion

A phytochemistry study of flavonoid composition [24] categorized the dahlia cultivars into three distinct types: (1) ivory-white cultivars containing flavones; (2) purple and pink cultivars with flavones and anthocyanins; and (3) red cultivars with flavones, anthocyanins, and chalcones. Interestingly, W of our bicolor cultivars did not align with the first group due to the absence of flavone accumulation, resulting in a pure-white appearance rather than an ivory-white phenotype. In contrast, R corresponded to group 3, characterized by the presence of flavones, anthocyanins, and chalcones. Moreover, the phenotypic instability of W of bicolor cultivars contrasts with the phenotypic stability observed in ivory-white cultivars. These differences suggest that additional fine-tuned regulatory mechanisms could contribute to the formation and maintenance of color patterns in dahlias.

Previous observation of the red turning resulting from feeding petals with a precursor (naringenin and taxifolin) demonstrated that the ABP downstream of the step of CHS remains active in the pure white dahlias [2]. In our study, PA accumulation in W indicated that the upstream ABP was not totally active despite the absence of anthocyanin accumulation. previous study found that all other EBGs and LBGs were active, except DvCHS1 and DvCHS2, suppressed by PTGS, as evidenced by sRNA read mapping [2]. Our findings showed the significant suppression of all upstream ABP genes, including PAL-C4H-4CL of the general phenylpropanoid pathway, indicating that the low flavonoid accumulation in W did not result from the stably expressed ANR. However, genome browser analysis of sRNA reads mapping to ABP genes did not reveal evidence of PTGS. Besides the upstream ABP genes, 3AT, which is responsible for acyl transfer to produce pelargonidin-3-(6-caffeoyl)G and cyandin-3-(6-caffeoyl)G in dahlias, was most significantly depressed. The significant suppression of CHS and normal expression of IVS align with previous observations of siRNA-induced DvCHS PTGS and stable DvIVS abundance [2,7]. Furthermore, in contrast to the normal ABP maintained by active ABP genes and regulatory TFs in the white petals of the bicolor dahlia ‘Yuino’ [2,7,8], the ABP genes in our study exhibited lower expression levels in W. This reduced abundance of ABP genes appears to be attributed to transcription factor repressors targeting multiple ABP genes. We propose that the lack of transcriptional activation, rather than PTGS, contributes to the depression of key genes such as 3AT in this study.

Anthocyanin biosynthesis is restricted to tissues containing functional MBW, where MYBs play a central role in regulating complex pigmentation patterns [14,15,16]. Among ABP activators, most showed similar expression levels, except for a suppressed SG6 R2R3-MYB (DpMYB1). Homologous to AN2 regulates distinct color patterns in Petunia through MBW [14,15,16]. DpMYB1 suppression could result in insufficient MBW formation, thereby failing to activate ABP genes. At the PPI level, a miR156-targeted DpSPL9, which was significantly induced in W, was identified via family analysis. Elevated levels of DpSPL9 may destabilize MBW complexes by competitively interacting with MYB components, acting as a predatory protein that hinders MBW function [19]. To sum up, we propose a putative anthocyanin synthesis model of the bicolor cultivar ‘LiRen’ (Figure 6). In the red petal bases, MBWs (DpMYB1-DpDEL/DpIVS-DpWRD1/DpWRD2) effectively activate ABP genes, facilitating the production of flavones, flavones, anthocyanins, and chalcones via EBGs like CHS or LBGs like GT. In contrast, in the white petal tips, suppressed DpMYB1 leads to reduced availability of MYB1 TF, weakening ABP activity. Additionally, miRNA regulation influences pigmentation. When miR156 suppression is relieved, DpSPL9 becomes active, further dismantling MBWs to weak ABP actives. The low abundance of GT in white tips inhibits the conversion of leucoanthocyanidins to anthocyanins, while PA is generated through the normal expression of ANS.

5. Conclusions

To investigate the formation mechanism of polychromatic petals, the activation of ABP and expression of candidate regulators were compared in red petal bases and white petal tips of bicolor dahlia based on RNA-seq and PA and anthocyanin detection. The lack of PA and the reduction in proanthocyanin of white petal tips resulted from the weaker APB genes. The depression of activation complex MBW due to the low abundance of MYB1 expression could be responsible for the lack of strong activation in white tips as well as the competitive binding to MYB1 via the miR156-SPL9 module. The model of miR156-SPL9 and MYB1 provided a candidate mechanism of polychromatic petal formation, and further genetic validation should be performed in the future.