Identification of Key Regulatory Genes Associated with Double-Petaled Phenotype in Lycoris longituba via Transcriptome Profiling

Abstract

Lycoris longituba produces a single flower bearing six tepals. The double-petaled phenotype of L. longituba has gained significant interest in China due to its ornamental and commercial value in tourism industries. This double-petal phenotype, characterized by stamen petalization, shows improved esthetic characteristics compared with conventional single-petal form. However, the molecular mechanisms underlying this floral trait remain largely undefined. In this study, RNA-based comparative transcriptomic analysis was performed between single- and double-petaled flowers of L. longituba at the fully opened flower stage. Approximately 13,848 differentially expressed genes (DEGs) were identified (6528 upregulated and 7320 downregulated genes). Functional annotation through Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses revealed several DEGs potentially involved in double-petal development. Six candidate genes, including the hub genes LlbHLH49, LlNAC1, LlSEP, LlTIFY, and LlAGL11, were identified based on DEG functional annotation and weighted gene co-expression network analysis (WGCNA). Transcription factors responsive to phytohormonal signaling were found to play a pivotal role in modulating double-petal development. Specifically, 123 DEGs were involved in phytohormone biosynthesis and signal transduction pathways, including those associated with auxin, cytokinin, gibberellin, ethylene, brassinosteroid, and jasmonic acid. Moreover, 521 transcription factors (TFs) were identified, including members of the MYB, WRKY, AP2/ERF, and MADS-box families. These results improve the current understanding of the genetic regulation of the double tepal trait in L. longituba and offer a base for future molecular breeding strategies to enhance ornamental characteristics.

1. Introduction

Petal morphology plays a crucial role in determining the ornamental value of flowering plants. Based on petal number, flowers are generally classified into single-, double-, and multi-petaled types [1]. A flower usually consists of four parts: sepals, petals, stamens, and pistils. Single-petaled flowers have only one layer of petals, and the petals are wide and flat. Double-petaled and multi-petaled flowers have two or more layers of petals. Numerous studies postulate that pistil and stamen petaloid are the most common way to form double-petaled flowers, and the degree of petaloid determines the appearance of flowers with different morphology. Various angiosperm species, such as Lagerstroemia speciosa [2], Dianthus chinensis [3], Jasminum sambac [4], Brassica napus [5], Eriobotrya japonica [6], Alcea rosea [7], and Prunus mume [8], show diverse floral morphologies largely due to the petaloid transformation of stamens. A double petal is one of the main ornamental traits of flowering organs in angiosperms. Double-petaled phenotypes are thus selected by breeders in many species because of the high ornamental value of double-petaled flowers.

The MADS-box transcription factor (TF) family represents a basic regulatory group involved in plant growth, development, and signal transduction processes [9]. Coen et al. (1991) classified floral organ identity-related MADS-box genes into distinct functional classes and proposed the ABC model of flower development, a widely accepted framework describing the genetic regulation of floral organogenesis, primarily regulated by members of the MADS-box gene family [10]. In this model, class A genes APETALA1 (AP1) and APETALA2 (AP2) specify sepal identity. The combined expression of class A (AP1, AP2) and class B genes, APETALA3 (AP3) and PISTILLATA (PI), determines petal formation. Class B (AP3, PI) and class C genes, particularly AGAMOUS (AG), jointly regulate stamen development, whereas class C genes (AG) alone are essential for pistil specification [7]. Further studies have expanded the model to include class D genes, such as SEEDSTICK (STK), and class E genes, known as SEPALLATA (SEP) [11,12]. The classical ABC model of floral organ development has been progressively improved into the more comprehensive ABCDE model [13]. In this extended model, classes A, B, C, and E genes collaboratively assemble into a tetrameric complex, often called the “tetramer model” [14]. Class-A genes (AP1 and AP2) in combination with class-E genes (SEP) regulate sepal development, while the development of petals is directed by class-A (AP1 and AP2), class-B (AP3 and PI), and class-E (SEP) genes. Pistil formation is regulated by class-C (AG) and class-E (SEP) genes. Class-D genes (e.g., STK), although critical for ovule development, exert minimal effect on the overall floral architecture and are thus frequently excluded from the floral patterning model, leading to its designation as the ABCE model [15]. The class A gene AP2 has been shown to influence double-petal formation in Brassica napus, Dianthus chinensis, Prunus persica, Rosa rugosa and Sagittaria sagittifolia [5,16,17,18,19,20]. Loss-of-function mutations in ABCE genes can induce homeotic transformations, substituting one floral organ type with another and leading to developmental anomalies [21]. Growing evidence has indicated that mutations or deletions in class-B and class-C genes are strongly associated with morphological variations in floral structures [22,23,24]. For example, in the Ranunculaceae family, downregulation or loss of AP3-3 gene expression has been closely associated with petal loss [22]. LMADS1, a member of the AP3 subfamily in Lilium longiflorum, shares high sequence similarity with other AP3 homologs. Ectopic expression of a truncated form of LMADS1 in Arabidopsis thaliana generated dominant-negative mutations, resulting in homeotic conversions of petals into sepal-like structures and stamens into carpel-like structures [23]. Therefore, there are differences in the function of the AP2 gene among different species. A mutation in the C-function AG gene leads to the homologous transformation from stamens to petals, and the number of stamens decreases, while the number of petals increases [25]. Currently, this has been similarly reported in many species, including Japanese morning glory (Pharbitis nil), Japanese mustard, Xanthoceras sorbifolium, Matthiola incana, Petunia hybrida and Kerria japonica, and Japanese rose (Rosa rugosa Thunb) [26,27,28,29,30,31,32,33]. The loss of function or abnormal expression of the C homologous gene often leads to the formation of double flowers.

In addition to genetic regulation, phytohormones play crucial roles in floral organ development by modulating gene expression through signal transduction and biosynthetic pathways [34,35,36]. Hormonal factors have been involved in increasing petal numbers and inducing double-petaled phenotypes. For example, gibberellic acid (GA₃) treatment in Impatiens balsamina has been shown to cause a transition from semi-double to fully double flowers [37], and exogenous application of auxin (indole-3-acetic acid, IAA) and GA₃ in Dianthus caryophyllus cultivar ‘White Sim’ was found to enhance petal numbers. Auxin influences the expression of AUX/IAA, ARF, and GH3 genes, therefore affecting cell division, elongation, and various physiological responses [34]. By regulating CRE1, AHP, and B-ARR genes, cytokinins promote cellular proliferation and differentiation [35]. Moreover, cis-acting regulatory elements within the promoter regions of MADS-box genes in Prunus campanulata ‘Plena’, suggest that multiple hormones—including auxin, abscisic acid, gibberellin, methyl jasmonate, and salicylic acid—may work together to influence the transcriptional regulation of genes involved in floral development [36,37]. Although multiple studies have been conducted on the role of plant hormones in seed germination, dormancy, floral bud differentiation, and general floral regulation [38], the specific hormone signaling pathways involved in stamen petaloid transformation remain insufficiently characterized. Besides their intrinsic signaling cascades, phytohormones may also affect the transcriptional activity of genes associated with floral organogenesis. However, the precise molecular mechanisms through which hormonal signals regulate gene networks responsible for floral organ identity, particularly in petaloid stamen formation, have yet to be fully elucidated.

The genus Lycoris Herbert, belonging to the family Amaryllidaceae, comprises approximately 20 species native to the moist, warm-temperate woodlands of East Asia [39]. Increasing horticultural interest in Lycoris has underscored its economic value, particularly in rural revitalization and ecotourism initiatives across China. In certain plant groups, the sepals and petals are not morphologically distinct and instead appear similar in form and color. In such cases, these undifferentiated, petaloid components are collectively referred to as tepals, as seen in families such as Magnoliaceae, Liliaceae, and Amaryllidaceae [40]. Based on the number of tepals, L. longituba can be categorized into single-petaled and double-petaled flowers. The ordinary single-petaled flower has only six tepals, but the double-petaled flower has more than eighteen tepals. The double-petaled phenotype is characterized by an increased number of tepals or additional whorls of tepals in a flower. Because of its ornamental value, this phenotype has been one of the major breeding objectives in floricultural plants, including rose, lilies, Dianthus caryophyllus, and chrysanthemum [16,28]. The development of novel L. longituba varieties with double-petal traits is expected to significantly enhance the species’ ornamental and commercial value. This study systematically examined morphological differences between the floral organs of single- and double-petaled L. longituba. Two representative cultivars were selected to assess the distinctions in floral organ primordia differentiation at transcriptomic levels. The primary objective was to determine whether the petaloid transformation of stamens in L. longituba is regulated by AP23 (Cluster-71145.0, Cluster-71145.2, and Cluster-52140.0) and AGL11 (Cluster-31235.0) in the ABCE floral development model, as well as to identify key genes associated with the regulation of double flowers and elucidate their regulatory networks. This study offers valuable insights into the anatomical and developmental differences underlying the single- and double-petaled phenotypes in L. longituba. It is expected to establish a foundational basis for elucidating the molecular mechanisms for the double-petal trait in this species.

2. Materials and Methods

2.1. Plant Materials

Lycoris longituba (L’Her.) Herb specimens were cultured at the Experimental Plantation of the Institute of Botany, Jiangsu Province, and the Chinese Academy of Sciences (Nanjing, China). Floral organ counts were performed using single-petaled and double- petaled cultivars collected in July 2024. According to the studies reported previously [41,42], three biological replicates of L. longituba flowers were sampled at stage 3 (fully opened flower), as shown in Figure 1A. For each cultivar, fifteen flowers were randomly sampled from different orientations and branches at the fully opened flower stage. Flower buds intended for transcriptome sequencing were collected from the Nanjing Botanical Garden, Memorial Sun Yat-Sen. Buds with uniform shape and size were collected daily in biological duplicates; one part was used for anatomical examination of developmental stages, while the other was immediately cryopreserved in liquid nitrogen and stored at -80 °C. Three biological replicates from single-petaled plants (SF-1, SF-2, SF-3) and three biological replicates from double-petaled cultivars (DF-1, DF-2, DF-3), with each replicate consisting of five flower buds.

2.2. RNA Extraction, Library Construction, Sequencing, and Transcriptome Assembly

RNA extraction, library construction, sequencing, and assembly Total RNA was extracted from all samples using the Plant RNA Extraction Kit (Tiangen, Hangzhou, China, DP441) following the manufacturer’s instructions. The quality and quantity of the RNA were examined by the Agilent 2100RNA Nano 6000 Assay Kit (Agilent Technologies, Santa Clara, CA, USA). Samples with RNA Integrity Number (RIN) ≥ 7 were subjected to cDNA library construction using the TruSeq Stranded mRNA LTSample Prep Kit (Illumina, San Diego, CA, USA). The resulting six cDNA libraries were pair-end sequenced on an Illumina NovaSeq 6000 platform (Illumina), which yielded 150 bp double ended sequencing reads. Raw sequencing data underwent quality control, including removing adapter-containing reads, reads with poly-N sequences, and low-quality sequences (more than 50% of bases with Qphred scores below the threshold). Transcriptome assembly was performed using Trinity2 software [43], and hierarchical clustering via Corset was employed to extract the longest cluster sequences for downstream analysis [44]. Transcript and cluster sequence lengths were statistically summarized. Coding sequences (CDSs) were predicted through alignment with the NR and Swiss-Prot protein databases. Open reading frames (ORFs) were identified for aligned transcripts, and nucleotide sequences were translated into amino acid sequences using standard codon tables (5′→3′). Transcripts without reliable alignments were analyzed for ORF prediction using Estscan (v3.0.3). The raw data was uploaded to the NCBI website with PRJNA1292443. Gene annotation was conducted by referencing seven databases: Nr (NCBI non-redundant proteins), Nt (NCBI non-redundant nucleotides), Pfam (protein families) (http://pfam.xfam.org/ (accessed on 20 November 2024)), KOG/COG (clusters of orthologous groups), Swiss-Prot (reviewed protein database), KEGG, and GO. Gene expression levels were quantified using fragments per kilobase of exon model per million mapped fragments (FPKM).

2.3. Differential Expression and Enrichment Analyses

Differential expression analysis was conducted to identify DEGs between (single- and double-petaled) plants. DEGs were determined via the DESeq2 R package with thresholds of a false discovery rate (FDR) ≤ 0.05 and absolute log2FC ≥ 1. The K-means algorithm was used to categorize these DEGs into several sets with similar expression patterns via the R package (version 3.6.3). GO and KEGG enrichment analyses of DEGs were performed using the cluster Profiler R package. Venn diagrams and expression heatmaps were plotted using R software. Pearson’s correlation coefficients between TF expression levels and differential metabolite content were calculated via R (v4.0.3), and correlation networks were visualized in Cytoscape (v3.8.2).

2.4. Weighted Gene Co-Expression Network Analysis (WGCNA)

WGCNA was conducted using the WGCNA R package, following a previously established methodology with modifications. Transcripts with FPKM > 10 were selected for network construction. Selection of the soft thresholding power was based on achieving a scale-free topology, with the minimum power chosen when the correlation coefficient exceeded or plateaued at 0.8. Pairwise Pearson correlation coefficients (PCCs) were calculated between all genes and converted into an adjacency matrix. Topological overlap (TO) was computed. Hierarchical clustering based on TO similarity was used to generate gene modules via the dynamic hybrid tree cut algorithm. Parameters included a minimum module size of 30 and a minimum module merging height of 0.20. Each module was represented by its module eigengene (ME). Module membership (kME) values were used to examine gene connectivity, and genes with kME > 0.7 were identified as hub genes. The content data of key genes were added as phenotypic traits for module–trait correlation analysis via PCC, with network visualization performed using Cytoscape (v3.9.1).

2.5. RNA Extraction, cDNA Synthesis, and qRT-PCR Assay

Total content of RNA was extracted using the RNAprep Pure Plant Kit (BC508, Huayueyang, Beijing, China) as per the manufacturer’s protocol. First-strand cDNA was synthesized using the PrimeScript™ II 1st Strand cDNA Synthesis Kit (TaKaRa Bio, Dalian, China). Next, qRT-PCR was performed using the SYBR^® Premix Ex Taq™ II kit (TaKaRa Bio, Dalian, China) on the Bio-Rad iQ5 Real-Time PCR System (Bio-Rad, Hercules, CA, USA). Each 15 μL PCR reaction mixture contained 5.9 μL ddH2O, 1 μL cDNA template, 0.3 μL each of forward and reverse primers (20 μM), and 7.5 μL of 2X TransStart^® Top Green qPCR SuperMix (Takara Bio, Dalian, China ). The PCR cycling conditions included initial denaturation at 95 °C for 5 min, followed by 40 cycles of 95 °C for 5 s and 60 °C for 30 s. The TIP41-like protein (TIP41) gene was used as an internal reference, and relative gene expression levels were calculated using the 2^−ΔΔCt method [39,45]. Primer sequences are listed in Table S1.

2.6. Statistical Analysis

All experimental data were repeated in three distinct biological replicates. Data was statistically analyzed using Student’s t-test as appropriate and described in the respective figure legends. Statistical significance was denoted by asterisks (* p < 0.05, ** p < 0.01). Alphabetic superscripts were used to represent statistically significant differences between treatment groups.

3. Results

3.1. Morphological and Genetic Characterization of Double-Petaled Phenotype in Lycoris longituba

The typical wild type of L. longituba shows a racemose inflorescence, is unisexual, and has a monoecious reproductive pattern. Flower morphology is essential in determining the ornamental value of this species. The structural features of single-petaled (SF) and double-petaled (DF) cultivars were compared to explore the morphological differences between floral types. The total number of floral organs was mostly consistent across both phenotypes (Figure 1). An initial analysis involved counting individual floral organs in single- and double-tepal types. As shown in Figure 1A, there were no statistically significant variances in the average counts of tepals and petaloid sepals between the two groups. However, petaloid modifications of tepals and stamens were observed in both cultivars. The average number of petaloid and total stamens was markedly elevated in the double-petaled cultivar compared to the single-tepal type. A remarkable difference (p < 0.01) in petal count was also recorded, with the single-petaled specimens averaging 6.00 ± 0.20 petals. In comparison, double-petaled flowers presented a significantly higher average of 18.28 ± 4.82 petals (Figure 1, Table S2). The total number of flowers was substantially higher in the double-petaled type, averaging 9.00 ± 0.90, relative to 6.00 ± 1.15 in the single-petaled form. This increase in floral organ quantity in the double-petaled genotype correlates with an enlargement of the floral meristem primordia.

3.2. Transcriptome Sequencing Data Analysis

To investigate gene expression differences related to the floral dimorphism in L. longituba, transcriptome sequencing was performed on mature flower buds collected at the tepal differentiation stage from single- and double-petaled cultivars. After filtering for quality to eliminate adaptors, poly-N stretches, and low-quality reads, approximately 45.42 Gb of high-quality sequencing data was obtained, with Q30 scores exceeding 94.88%. More than 98.52% of bases met a Q20 threshold, and over 95.26% surpassed the Q30 benchmark (Table S3). The GC content of clean reads was calculated at 44.64%, indicating reliable sequencing quality. These results confirmed the high quality of the transcriptomic data. De novo transcriptome assembly produced 240,461 transcripts and 131,619 unigenes. Among these, 22.76% of unigenes ranged from 300 to 500 base pairs, while 23.80% of transcripts were between 500 and 1000 bp. Moreover, 40.25% (75,820) of unigenes exceeded 1000 bp in length (Figure 2A). Species distribution analysis of annotated sequences showed that most shared homology with Asparagus officinalis (52.29%), followed by Elaeis guineensis, Phoenix dactylifera, Cocos nucifera, Ensete ventricosum, Ananas comosus, and Dioscorea alata (Figure 2B), reflecting close evolutionary relationships. In this analysis, approximately 186,338 CDSs were identified, with a broad size distribution from 201 nucleotides (nt) to 14,127 nt (Figure 2A, Table S4). The majority of CDSs ranged from 400 to 3000 nt. Furthermore, 3663 TFs were annotated, representing 59 gene families (Table S5, Figure S1). Among these, the basic helix–loop–helix (bHLH) family was the most abundant, with 186 members, followed by the NAC family with 162 members, the MYB-related group with 159 members, and the AP2/ERF-ERF family containing 158 members.

GO-based classification of the annotated unigenes from the transcriptome database revealed enrichment across three main categories: biological process (BP), cellular component (CC), and molecular function (MF) (Figure 2C). Within the BP category, the term “cellular process” was the most frequently represented. For CC, “cellular anatomical entity” showed the highest level of enrichment, while in the MF category, “binding” appeared as the dominant function. Sample correlation analysis also confirmed high consistency among biological replicates within each experimental condition (Figure 2D). Further functional classification using the KOG database indicated that the most enriched categories included “general function prediction only,” “posttranslational modification, protein turnover, chaperones,” and “signal transduction mechanisms,” corresponding to 9998, 5104, and 4422 genes, respectively (Figure S2).

To ensure that the replicates behaved similarly and to examine the overall trends in the data, principal component analysis (PCA) and Pearson’s correlation coefficient were performed. These findings revealed that the double-flower sample induced a shift in the lavender transcriptome (Figure 2D and Figure S3). The RNA-seq dataset was thus considered suitable for in-depth transcriptomic analysis. Approximately 13,848 DEGs were identified between single- and double-petaled L. longituba at the floral bud stage, comprising 6528 upregulated and 7320 downregulated genes (Figure 3A). Comparative DEG analysis between DF and SF stages revealed 27,952 DEGs in DF-1 vs. SF-1, 21,916 in DF-2 vs. SF-2, and 17,602 in DF-3 vs. SF-3 (Figure 3A, Table S6). Generally, more upregulated genes were observed than downregulated ones, except in the DF-3 vs. SF-3 contrast. This pattern indicates a high transcriptional activity related to floral differentiation in L. longituba. Analysis of transcript isoform variation showed that 6860 DEGs were shared across all DF vs. SF comparisons. However, 13,211, 6193, and 1974 DEGs were uniquely detected in DF-1, DF-2, and DF-3 stages, respectively (Figure 3B and Figure S4). These results indicate that numerous genes may be activated or suppressed at specific stages of flowering.

3.3. Functional Enrichment and Differential Transcript Expression Analysis

Differentially expressed genes (DEGs) carried out multiple functions in identifying key genes associated with the regulation of double flowers and elucidating their regulatory networks in L. longituba, and the DEGs were analyzed through GO enrichment analysis for biological process, cellular composition, and molecular function. The top 50 most significantly enriched GO terms were grouped into three main categories: molecular function (MF), cellular component (CC), and biological process (BP). The results showed that most DEGs were enriched in the BP category, especially those involved in the apical plasma membrane (GO:0016324), pectin catabolic process (GO:0045490), regulation of pollen tube growth (GO:0080092), and modulation of cell shape during differentiation (GO:0010769). In the CC category, genes were mainly associated with the apical part of the cell (GO:0045177) and membrane-anchored components (GO:0031225). In the MF category, significant enrichment was found in enzymatic activities such as pectinesterase inhibitor activity (GO:0046910), terpene synthase activity (GO:0010333), and aspartyl esterase activity (GO:0045330) (Figure 3C). Next, KEGG pathway enrichment analysis was conducted to identify key metabolic processes associated with the DEGs. Significant enrichment was observed in 20 metabolic pathways, especially those involved in pentose and glucuronate interconversions, biosynthesis of phenylalanine, tyrosine, and tryptophan, ascorbate and aldarate metabolism, and C5-branched dibasic acid metabolism. Among these, flavonoid biosynthesis and isoquinoline alkaloid biosynthesis pathways were associated with secondary metabolite production (Figure 3D). These pathways, particularly those related to amino sugar and nucleotide sugar metabolism, were found to play vital roles in regulating petal number and size in L. longituba. To investigate the difference between SF and DF on gene expression, all DEGs were analyzed by Mfuzz and grouped into six clusters, containing 4300, 7587, 2706, 11,798, 6345, and 9539 genes, respectively (Figure 4A). Genes from cluster 6 were downregulated between double-petaled and single-petaled flowers, while genes from clusters 1 and 5 were upregulated between single- and double-petaled L. longituba at the floral bud stage; therefore, genes from these three clusters were identified as responsive genes (Figure 4). To identify core regulatory genes associated with specific stages of floral development, particularly in comparison to the closed bud phase of L. longituba, KEGG pathway enrichment was conducted for DEGs within Clusters 1, 5, and 6. Clusters 5 and 6 were further examined due to the relevance of their gene functional annotations. Among the top 20 enriched pathways in these clusters, “cutin, suberine and wax biosynthesis,” “MAPK signaling pathway,” “pentose and glucuronate interconversions,” and “plant hormone signal transduction” emerged as key metabolic processes potentially regulating petal size development in L. longituba (Figure 4B–D). The majority of the biological processes mentioned above are associated with the regulation pathway and mechanism of double flower formation in L. longituba.

3.4. Transcription Factors and WGCNA

Transcription factors (TFs) are vital regulators of plant growth and development. This study identified approximately 552 differentially expressed TFs belonging to 69 families (Table S7). The most common families included MYB (39 members), AP2/ERF-ERF (35), WRKY (28), NAC (25), GARP-G2-like (22), bZIP (21), bHLH (20), and C2H2 (18). These TFs participate in various developmental and hormonal processes, including floral organ formation. To explore transcriptional regulation, a WGCNA was performed, revealing 30 distinct co-expression modules, each represented by a specific color (Figure 5). Many TFs clustered in the blue and turquoise modules, including 51 MYBs, 48 WRKYs, 45 AP2/ERFs, and 15 MADS-box genes (Figure 6A). Expression analysis of WRKY, AP2/ERF, and MADS TFs showed two different expression patterns (Figure 6B–D). Transcripts such as AP2 (Cluster-71145.0, Cluster-71145.2, and Cluster-52140.0) and ERF105 (Cluster-66083.3) were significantly upregulated in double-petaled cultivars, while other AP family members showed decreased expression. Furthermore, the differential expression of MADS and WRKY TFs across the blue and turquoise modules reflected their opposing regulatory roles. Several key MADS-box genes, including AGL20 (Cluster-912.1, Cluster-65151.12, and Cluster-65151.6), AGL62 (Cluster-19045.0), AGL61 (Cluster-63155.2), AGL104 (Cluster-4775.0), AGL104 (Cluster-4775.1), AGL20 (Cluster-65151.9), AGL30 (Cluster-45098.0), and AGL11 (Cluster-31235.0) showed higher expression in single-flowered phenotypes (Figure 6C). These findings suggest that variation in petal number in L. longituba may be driven by changes in the expression of WRKY, AP2/ERF, and MADS-box TFs or their downstream targets involved in floral organ development.

Among these, three distinct co-expression modules (MEblue and MEturquoise) were identified, comprising 6259 and 9099 isoforms, respectively. These modules showed strong associations with petal number variation, as indicated by the expression patterns of their respective module eigengenes between single-petaled and double-petaled L. longituba (Figure 7). Therefore, isoforms within the MEblue and MEturquoise modules were analyzed using KEGG pathway enrichment and network analyses (Figure 7A,C, Tables S8 and S9). Isoforms in the MEblue module were mainly enriched in metabolic pathways like carbon metabolism, glyoxylate and dicarboxylate metabolism, selenocompound metabolism, the tricarboxylic acid (TCA) cycle, amino acid biosynthesis, and tryptophan metabolism. However, the MEturquoise module was substantially enriched in pathways related to photosynthesis antenna proteins and plant hormone signal transduction. Both modules showed significant enrichment in the MAPK signaling pathway, suggesting their roles in environmental signal processing. These findings indicate that these biological processes, particularly those involved in carbohydrate metabolism, light-dependent energy conversion, and hormonal signaling, may play key roles in regulating petal number determination and floral development in L. longituba.

Two specific co-expression modules, MEblue and MEturquoise, showed strong positive correlations with floral development in L. longituba (r ≥ 0.91) (Figure 7B,D). Hub genes within these modules were identified using strict selection criteria: a module kME ≥ 0.9 and an edge weight ≥ 0.5 (Tables S10 and S11). These hub genes displayed extensive interconnectivity with neighboring genes, occupying central positions within their respective networks and serving as key components of the co-expression structure. About 150 genes in both MEblue and MEturquoise modules exceeded the defined kME threshold, indicating strong intramodular relationships and maintaining the integrity of the co-expression network. Among these, several TFs involved in aroma compound biosynthesis were identified as core regulatory elements. Five TFs were found within the MEblue module and two within the MEturquoise module. These included LlSEP (Cluster-48411.4), LlTify (Cluster-44311.1), LlNAC1 (Cluster-22839.2), LlbHLH49 (Cluster-39352.0), LlSNF2 (Cluster-22873.17), LlAG (Cluster-31235.0), and RWP-RK (Cluster-2940.3) (Figure 7). Based on the results of the WGCNA, we hypothesize that the aforementioned genes are associated with floral organ formation.

3.5. Phytohormone-Mediated Regulation of Double-Flower Formation

Phytohormones play essential roles in regulating various processes of plant growth and development. In our research, KEGG pathway enrichment was conducted for DEGs within Clusters 6, and MEturquoise module were mainly enriched in plant hormone signal transduction pathways (Figure 4 and Figure 7). Given that the petal is the key and showy floral organ of ornamental plants, the effects of phytohormones on petal growth has garnered considerable attention. Several classes of phytohormones, such as auxin, abscisic acid (ABA), brassinosteroids (BR), cytokinins (CTK), ethylene, gibberellins (GA), and jasmonic acid (JA), are involved in controlling petal size through complex interactions that affect both cell proliferation and expansion (Figure 8). Transcriptome analysis in this study identified 123 DEGs linked to hormone biosynthesis and signal transduction pathways (ko04075), covering multiple distinct hormonal pathways (Figure 8). In the cytokinin (CTK) signaling cascade, four Arabidopsis histidine-containing phosphotransfer (AHP) (Cluster-60675.0, Cluster-60675.10, Cluster-60675.11, and Cluster-60675.6) proteins and one B-ARR gene (Cluster-51685.0, Cluster-54604.7, Cluster-57371.1, Cluster-57904.3, and Cluster-63330.4), were upregulated in DF compared to SF. They regulate cell division and shoot initiation. In the JA signaling pathway, six genes were upregulated in DF, all of which are involved in α-linolenic acid metabolism. These included three Jasmonate ZIM-domain (JAZ) proteins (Cluster-41993.0, Cluster-44311.1, and Cluster-61692.5) and three MYC2 TFs (Cluster-47020.0, Cluster-65262.2, and Cluster-71657.4), which are integral to JA-mediated responses and the regulation of floral organ development. Moreover, ten DEGs involved in the ethylene signaling pathway, particularly in cysteine and methionine metabolism, were identified. These included an ethylene-resistant gene (ETR, Cluster-18103.7), two constitutive triple response 1 (CTR1, Cluster-23752.8 and Cluster-9961.0), and an ethylene-insensitive 2 (EIN2, Cluster-31327.7), all of which showed increased expression in DF relative to SF (Figure 8). AUX/IAA is a repressor of ARFs and undergoes proteasome-mediated degradation after auxin signal perception. Together with small auxin-up RNA (SAUR) and the auxin-responsive GH3 family, these three gene families are the core components of the early auxin signaling cascade and regulate cell expansion during various stages of plant development. In the current study, 31 DEGs associated with auxin signaling were identified. These included four AUX genes (Cluster-55087.1, Cluster-55087.10, Cluster-65002.6, and Cluster-68327.8), one TIR gene (Cluster-17531.1), one AUX/IAA gene (Cluster-31394.1), four ARF genes (Cluster-22656.7, Cluster-59725.4, Cluster-62662.0, and Cluster-64709.1) and seven SAUR gene (Cluster-37806.1, Cluster-57315.68, Cluster-57315.69 Cluster-60347.7, Cluster-71054.5, Cluster-71829.2, and Cluster-71829.5). Furthermore, genes involved in GA signaling were also found to be differentially expressed. Two GA-insensitive dwarf1 (GID1) (Cluster-34482.3 and Cluster-34482.5) genes and three DELLA genes (Cluster-47468.5, Cluster-68528.1, and Cluster-68528.2), which participate in diterpenoid biosynthesis and regulate stem elongation and germination, were upregulated in DF samples relative to SF (Figure 8). Brassinosteroid (BR) biosynthesis-related genes also demonstrated altered expression patterns. Approximately 19 DEGs involved in BR signaling were found to be upregulated in DF samples. These included four brassinosteroid insensitive (BIN2) genes (Cluster-43474.1, Cluster-43474.2, Cluster-43474.7, and Cluster-50272.2), one BZR1/2 gene (Cluster-29997.5), six BRI1 genes (Cluster-25640.66, Cluster-30423.2, Cluster-30423.3, Cluster-54310.2, Cluster-54505.0, and Cluster-69851.1), three BRI1-associated receptor kinase1 (BAK1) (Cluster-46412.14, Cluster-64528.6, and Cluster-68420.8), two brassinosteroid signaling kinase (BSK) (DN36014 and DN32815), three brassinosteroid insensitive2 (BIN2) (DN34098, DN22849-1, and DN22849-2), and five TCH4 genes (Cluster-57001.0, Cluster-57001.2, Cluster-57001.3, Cluster-57001.5, and Cluster-63592.1). These genes are associated with cell elongation and division, which are integral to floral organ development. The expression patterns of these hormone-related DEGs provide further validation for the transcriptomic findings in flower buds. The changes in hormone levels during flower bud development of L. longituba should be detected in the next research and underscored the pivotal role of phytohormonal regulation in the formation of double petals in L. longituba.

3.6. Validation of RNA-Seq Data by qRT-PCR

To confirm the accuracy and reliability of the transcriptomic data, 15 representative DEGs were selected for qRT-PCR analysis. The validation included genes showing up- and downregulated expression profiles in DF compared to SF. The upregulated genes in DF included LlSEP (Cluster-48411.4), LlTify (Cluster-44311.1), LlNAC1 (Cluster-22839.2), LlbHLH49 (Cluster-39352.0), LlSNF2 (Cluster-22873.17), LlMYC (Cluster-47020.0 and Cluster-65262.2), and LlAP23 (Cluster-65633.10). Conversely, the downregulated genes included LlWRKY40 (Cluster-70541.1), LlWRKY71 (Cluster-56961.1), LlAGL62 (Cluster-19045.0), LlARF (Cluster-34373.0), LlAGL20 (Cluster-65151.9), LlAGL30 (Cluster-45098.0), LlAGL11 (Cluster-31235.0). The qRT-PCR expression trends for 15 genes were consistent with those observed in the RNA-seq dataset, thus validating the transcriptomic analysis of the identified DEGs between SF and DF in L. longituba (Figure 9).

4. Discussion

4.1. Illumina RNA-Seq Provided Transcriptome Information for L. longituba

L. longituba is a widely cultivated ornamental species known for its vibrant flower coloration and variety of flower shapes [39]. Among its cultivars, the double-petaled variant is especially valued for its attractive structure and has become increasingly important in horticultural breeding. Despite its value in gardening, the molecular mechanisms behind the transition from stamens to petaloid structures are still not fully understood. In this study, high-throughput Illumina-based RNA sequencing was used to explore the transcriptomic profile related to floral development, focusing on the genetic factors involved in stamen petaloid formation. Comparing single-petaled and double-petaled types revealed key regulatory elements and provided a foundation for understanding the molecular processes that lead to double-petaled development.

Transcriptomic analysis provides a comprehensive platform for understanding genome-wide expression patterns and examining the molecular mechanisms of developmental processes. This study marked the first systematic transcriptome profiling of L. longituba floral buds at different stages of morphological differentiation. Differential gene expression analysis showed that genes involved in microtubule assembly and flavonoid biosynthesis pathways were significantly changed between floral types. The microtubule network has been linked to controlling tepal anisotropy, helping with tepal elongation and curvature [27]. Simultaneously, flavonoids have versatile roles in regulating auxin transport and signaling, and also act as mediators in hormonal crosstalk, transcriptional regulation, and oxidative stress responses [46]. The ongoing differential expression of genes in these pathways suggests a key role in the petaloid transformation observed in the double-petaled phenotype (Figure 3).

4.2. Differentially Expressed Genes Highlight Molecular Pathways Associated with Double-Tepal Formation

TFs are a crucial layer of transcriptional regulation in plant development, orchestrating complex regulatory networks by affecting downstream gene expression [47]. A total of 3662 TFs from 69 different families were identified from the transcriptome assembly. The most prevalent families include bHLH, MYB, NAC, B3, C2H2, bZIP, AP2/ERF, MADS-box, and WRKY (Figure 5A). A total of 521 DEG TFs were identified as involved in floral organ development, with a majority upregulated in double-petaled samples (Table S7). Several TFs identified in this study, particularly from the MYB, bHLH (LbHLH49), WRKY, NAC (LlNAC1), and GATA families, have previously been involved in petaloid development, supporting their potential functional relevance in L. longituba. For example, MYB TFs have been shown to physically interact with bHLH proteins, forming MYB-bHLH transcriptional complexes that regulate floral organogenesis and stamen maturation [48]. NAC domain TFs, known as ethylene-responsive, participate in petal development by regulating cell expansion and morphogenesis [49]. Additionally, significant hub genes of LlNAC1 (Cluster-22839.2) play a regulatory role in promoting petaloid transformation, and could be critically modulated following synthetic genetic circuit enabled programming to increase tepel number in L. longituba. Similarly, WRKY TFs have been involved in floral primordia differentiation, stamen identity, and stress-related signaling pathways [50,51]. In addition to the families above, other TFs, such as C2H2-type ZFPs, were also found to be differentially expressed. These TFs regulate flowering initiation, floral organ development, and pollen–pistil interaction [49]. In Brassica rapa, genes such as BrZFP244 and BrZFP187 show tissue-specific expression during stamen and pistil development, respectively [52]. TCP TFs also play significant roles in floral symmetry, timing, and jasmonate signaling, with CYC-like TCP members being important for inflorescence complexity [53,54]. Intriguingly, LlbHLH49 and LlNAC1 in L. longituba were significantly upregulated in double petals, suggesting their potential interaction with other factors to promote stamen-to-petal transformation (Figure 9). These TFs, in concert with hormonal pathways, form an intricate network shaping the unique floral architecture of L. longituba.

The ABCE model acts as a basic model for understanding floral organ identity specification, although its level of conservation varies considerably across different plant taxa. Within this regulatory system, class A genes, especially those coding for APETALA2 (AP2)-like TF, are crucial for regulating developmental phase transitions in plants [55]. In A. thaliana, the AP2-like gene family includes six members: AP2, TOE1 (TARGET OF EAT1), TOE2, TOE3, SMZ (SCHLAFMUTZE), and SNZ (SCHNARCHZAPFEN) [56]. Functional studies have shown that AP2 and TOE3 are essential for maintaining cellular identity and determining floral organ fate, while TOE1, TOE2, SMZ, and SNZ mainly regulate flowering time [57,58]. In this study, members of the AP2 family, specifically Cluster-71145.0, Cluster-71145.2, and Cluster-52140.0, were identified as DEGs that showed substantial upregulation in double-tepal L. longituba. This finding suggests that AP2 genes may play a regulatory role in promoting petaloid transformation (Figure 6B). Previous research has demonstrated a strong link between the expression levels of AGAMOUS (AG)-like genes, classified as class C, and the degree of stamen-to-petal homeotic conversion. In A. thaliana, the loss of AG function results in the transformation of stamens into petals [59]. Similarly, in Lilium brownii, decreased expression of LelAG1 during the development of the third and fourth floral whorls was associated with increased petaloidy [60]. Excessive accumulation of AP2 transcripts has been shown to suppress AG expression, leading to the abnormal transformation of stamens into petaloid structures [61]. Mutations in AG homologs have developed semi-double flower development in Dianthus caryophyllus [62]. Moreover, DcaAG genes reduce petal number but still play essential roles in defining stamens and carpels [3]. Similar regulatory patterns have been observed in Paeonia suffruticosa [63]. This study identified LlAG (Cluster-31235.0), an AG-like MADS-box gene, and its homologs through combined transcriptome sequencing analysis. These genes contained conserved MADS and K-box domains, indicating their functional integrity (Figure 6C). This expression pattern matched previous studies in A. thaliana, Prunus persica, and Rosa rugosa [64,65]. This mechanism has also been supported in several plant species, including Rosa spp. [66], ranunculids [67], and Lilium longiflorum [68], where lower AG transcript levels in the third floral whorl were associated with the replacement of stamens by petaloid structures [69]. Furthermore, qRT-PCR analysis confirmed the expression profiles of AGL11 (Cluster-31235.0) in both single- and double-petaled L. longituba (Figure 9). The results showed that AG expression was substantially lower in double flowers compared to single ones. According to the ABCE model of floral organ identity, the suppression of AG homolog expression has been related to the homeotic transformation of stamens into tepals, which contributes to the development of double-petaled phenotypes in L. longituba.

4.3. Plant Hormones Regulate Stamen Petalization

Previous studies have revealed that the development of floral organs, (viz. sepal, petal, stamen, and carpel) is to a large extent controlled by hormones, including auxin, gibberellins (GA), cytokinin (CK), ethylene, abscisic acid (ABA), jasmonic acid (JA), salicylic acid (SA), and brassinosteroids (BR) [70]. In this study, a total of 123 DEGs related to hormone signaling were identified as associated with stamen petaloid development in L. longituba. Previous studies highlighted the central role of auxin in stamen development, a multi-phase process involving both early organ formation and later stages such as pollen maturation, filament elongation, and anther dehiscence [71]. The YUCCA (YUC) gene family, which encodes flavin monooxygenases, is a key part of the auxin biosynthesis pathway [72]. Auxins were found to accumulate at the tip of stamen primordia in a polarized manner, affecting organ differentiation [71]. Genes encoding auxin/indole-3-acetic acid (AUX/IAA) and auxin response factors (ARFs) serve as downstream targets of auxin signaling and are crucial components of the nuclear auxin response pathway [73,74,75]. In A. thaliana, auxin signaling acts through receptors like TIR1 (Transport Inhibitor Response 1) and AFB (Auxin Signaling F-box proteins), which detect auxin molecules, promote degradation of AUX/IAA proteins via the ubiquitin–proteasome system, and free ARFs to activate auxin-responsive gene expression [76]. This signaling pathway appears to be conserved across many plant species, and disruptions in these components were reported to disrupt normal stamen development. Several ARFs were found to play crucial roles in floral development. For example, AtARF1 and AtARF2 affect floral meristem initiation, stamen development, and floral organ aging in A. thaliana [77]. AtARF mutants displayed changes in both stamen and petal numbers [78], while increased expression of PgARF in Punica granatum was related to transforming stamens into petaloid structures [79]. Similarly, in Rosa rugosa, the auxin-responsive transcription factor RhARF18 acted as a repressor of the class-C gene RhAG, regulating the stamen-to-petal transition through an auxin-dependent process [80]. Furthermore, higher expression of AtARF17 in A. thaliana was connected to abnormal floral development, smaller petals, and defective stamen formation due to misregulation of GH3-like genes [81]. In addition to auxin, several other classes of phytohormones are also closely related to stamen development. In the current study, the expression levels of IAA, ARF, TIR, SAUR, and AUX related to IAA biosynthesis and signal transduction were significantly higher in double tepals than those in single tepal. However, how AUX response factors interact with each other in L. longituba to affect the formation of double tepals are still unclear.

The occurrence and putative functions of JA and its derivatives, known as jasmonate(s), in floral development have been well reviewed in the last decade. JA perception is mediated by the COI1 receptor, which promotes the degradation of JAZ repressor proteins and then activates downstream TFs important for stamen development. Mutations in key parts of the JA signaling pathway caused problems like impaired filament elongation, delayed anther dehiscence, and lower pollen viability [82,83]. In A. thaliana, JA primarily affects the later stages of floral organ maturity, including stamen differentiation and petal expansion, and exogenous application of JA can partially rescue the defective floral phenotypes in related mutants [84]. The functional effects of JA on floral development are closely associated with the regulation of LOX and MYC genes, which mediate JA signaling and downstream responses. Changes in the expression of these genes have been reported to result in abnormal floral structures, including increased petaloidy [30]. In this study, the expression levels of JAZ and MYC genes involved in JA signal transduction were significantly lower in double flowers compared to single flowers, suggesting a possible suppression of JA signaling during the development of stamen petalization (Figure 8). Furthermore, interactions between JA and auxin signaling pathways are proposed to synergistically influence floral organ development. Crosstalk between these hormonal pathways may regulate the differentiation and identity of stamens and pollen fertility. The current transcriptomic results indicate that auxin and JA signaling components show different expression patterns related to petaloid formation in L. longituba. However, the exact molecular mechanisms controlling their interaction and the precise regulatory cascades involved in floral organ transformation still need to be clarified. These observations collectively indicate that auxin and JA are crucial in the molecular regulation of the transformation from stamen to petal in L. longituba, possibly through gene expression regulation related to hormonal synthesis, transport, and response pathways. Collectively, these findings highlight the complex interplay of JA signaling and biosynthesis genes in shaping the double-petaled phenotype of L. longituba.

Gibberellin (GA) signaling is mediated through the GA receptor GID1, which perceives gibberellin and recruits DELLA proteins [85]. This interaction targets DELLA repressors for degradation via the 26S proteasome pathway, activating downstream gene expression in GA responses. This signaling mechanism has considerably affected stamen development. Loss-of-function mutations in GID1a, GID1b, and GID1c resulted in defects in filament elongation and arrested anther development [76]. Gibberellins have been reported to promote stamen maturation by inducing the expression of MYB TFs, a process also modulated by JA [86]. In the present study, the expression levels of GID genes (Cluster-34482.3 and Cluster-34482.5) and DELLA genes (Cluster-47468.5), which are important for GA signal perception and transduction, were found to be significantly elevated in double petals relative to single petals, suggesting a potential role of gibberellin signaling in petaloid stamen development in L. longituba (Figure 8). Ethylene signaling was also involved in floral organ development, particularly by activating EIN2 and EIN3 TFs, mediating downstream responses via the EIN2–EIN3/EIL1 pathway. Activation of this pathway suppresses anther development [87]. In this study, EIN2/3 (Cluster-65795.0, Cluster-65795.2, Cluster-2413.0, and Cluster-3119.0) related to ethylene signal transduction and response in double tepals were significantly lower than those in single tepals, indicating a possible suppression of ethylene-mediated inhibition of stamen development. Furthermore, signaling pathways involving GA, ABA, salicylic acid (SA), and methyl jasmonate (MeJA) have been shown to interact with cis-regulatory elements within the promoter regions of MADS-box genes, thus influencing the development of stamens and petals [88]. External ethylene application has also been shown to induce stamen shortening or the conversion of stamens into pistil-like structures in certain plant species, emphasizing its key role in determining floral organ identity [89]. These results indicate that the hormonal regulation of tepal development in L. longituba involves a complex interaction between multiple signaling pathways. GA and ethylene exert opposing regulatory effects during the transition from stamens to petaloid structures. Therefore, the extent of hormone cross-regulation and conservation of floral organ regulation across plant species of varying morphologies are intriguing areas of research. Taken together, molecular regulation of floral organ numbers, especially those of petals, involves the integration of many genetic and environmental factors in L. longituba. Our discovery of this mechanism provides insight into the stage-specific regulation of flower development in L. longituba and, more broadly, in the Amaryllidaceae family, especially the formation of double petals. The function and regulatory mechanism of the key genes of this hormone model will continue to be further developed in follow-up research.

Genome variation in Lycoris includes chromosome numbersfrom 12 to 33 and genome sizes variying from 18.03 Gb to 32.62 Gb [90]. However, the widespread morphological variation and high diversity of Lycoris have created challenges in germplasm sorting, development, and utilization. These transcriptomic resources provide foundational insights into the molecular mechanisms of double-flower formation and have significant theoretical and practical implications for future genetic improvement efforts. The findings are expected to support the development of new cultivars with desirable floral traits through molecular breeding and germplasm innovation in Lycoris. Taken together, this study provides fresh perspectives on the regulatory processes governing double-flower formation in Lycoris flowers, and it proves valuable for the genetic engineering of desirable floral traits in Lycoris and other decorative plants.

5. Conclusions

This study concluded that the development of stamen petaloid structures is a highly complex biological process controlled by multiple regulatory networks. Transcriptomic profiling of L. longituba floral tissues identified 13,848 DEGs (6528 upregulated and 7320 downregulated) associated with the double-petaled phenotype. Among these, 123 DEGs were involved in plant hormone signaling pathways potentially contributing to floral organ development, while 552 DEGs encoded transcription factors related to regulating stamen petaloidy. It is proposed that the transformation of stamens into petaloid structures in L. longituba is primarily controlled by AP2 (Cluster-71145.0, Cluster-71145.2, and Cluster-52140.0) and AGL11 (Cluster-31235.0) in the ABCE model, along with additional transcription factors and hormone-responsive genes. These transcriptomic resources provide foundational insights into the molecular mechanisms of double-flower formation and have significant theoretical and practical implications for future genetic improvement efforts.