Introduction

Lilies are important ornamental plants that belong to the genus Lilium, which contains abundant species and cultivars. Commercial potted and cut lily cultivars are predominantly vegetatively-propagated (bulbs, bulb divisions and aerial axillary stem bulbils) rather than sexually (seed) (Beattie and White 1993; Griffiths 1934), and the seedlings of most lily species require a long vegetative growth stage of more than 1 year before flowering.

Seed-propagated white trumpet lilies, Lilium × formolongi, are important commercial cut flowers that are obtained from repeated interspecific crosses of Lilium formosanum Wallace × Lilium longiflorum (Okazaki 1994). This lily hybrid possesses the characteristic of flowering within a year after sowing (200–240 days) without vernalization, but is sensitive to photoperiod (Anderson et al. 2012; Sakamoto 2005). These characteristics are important to the development of seed-propagated, non-vernalization-requiring colored lily hybrids (Optiz et al. 2009). Because of the desirable characteristics of L. × formolongi, most reported research focuses on the cultivation and cross-breeding of this plant (asl Hamid and Kim 2011; Horita et al. 2003). However, the molecular regulatory mechanism of flowering is largely unknown, as is the genetic mechanism of blossoming in 1 year of L. × formolongi.

Studies on the biogenetics of flowering time in the model plant Arabidopsis thaliana have shown that the flower development period can be divided into several stages, with the first being flowering induction, in which the shoot apical meristem (SAM) remains in a vegetative state (Meyerowitz et al. 1991). In this process, the plant responds to internal and environmental signals, including the age of the plant, its nutritional state, day length and temperature. These physiological and genetic activities determine the change of SAM from vegetative to floral (Meyerowitz et al. 1991). According to our previous research, the seedlings of L. × formolongi are sensitive to the long day (LD) flowering pathway. In addition, the characteristic of prompt blossoming in 1 year from seeds indicates that the age-dependent and carbohydrate nutritional flowering pathways may also play important roles in the flowering initiation of L. × formolongi.

Genetic analyses of A. thaliana, a facultative long-day plant, have revealed many photoperiodic genes, including CRYPTOCHROME 1 (CRY1), CONSTANS (CO), GIGANTEA (GI), FLOWERING LOCUS T (FT) and FLOWERING LOCUS D (FD) (Turck et al. 2008). The circadian rhythm expression pattern of CO/FT is crucial in triggering flowering induction (Suárez-López et al. 2001; Meng et al. 2011). CO is a member of the CONSTANS-LIKE (COL) gene family, whose members encode putative zinc finger transcription factors (TFs) (Putterill et al. 1995). CO transcripts are regulated by the circadian clock and are diurnally regulated with a peak at night under both LDs and short days (SDs) (Andrés and Coupland 2012; Suárez-López et al. 2001). However, there is an additional peak of CO mRNA that occurs 12–16 h after dawn (in the light) under LD conditions, which is essential for the day-length-dependent promotion of flowering (Suárez-López et al. 2001). The CO protein in the dark is degraded by an ubiquitin ligase complex. Thus, only the peak of CO mRNA that occurs in the light at the end of a LD leads to CO protein accumulation (Andrés and Coupland 2012; Suárez-López et al. 2001). Subsequently, the accumulation of CO protein drives transcript accumulation of FT mRNA at the end of day under LD conditions, but not SD conditions, and triggers the flowering induction under LDs (Andrés and Coupland 2012; Kardailsky et al. 1999).

Recently, a new age-dependent flowering pathway, which is regulated by two microRNAs (miR156, miR172) and their target genes SQUAMOSA PROMOTER BINDING PROTEIN-LIKE (SPL) and APETALA2 (AP2)-like, was revealed in model plants (Bergonzi et al. 2013; Khan et al. 2014). Furthermore, sugar is an endogenous signal for the vegetative to reproductive stage transition in plants, and trehalose-6-phosphate (T6P) has been suggested to function as a proxy for carbohydrate status (O’Hara et al. 2013; Proveniers 2013). A recent study revealed that the TREHALOSE-6-PHOSPHATE SYNTHASE 1 (TPS1) gene catalyzes the formation of T6P and the loss of TPS1 causes A. thaliana to flower extremely late, even under inductive environmental conditions (Wahl et al. 2013).

The genome size of lily is very large, at ~36 Gb, which is ~300 times that of the A. thaliana genome (Du et al. 2015). Whole transcriptome sequencing using next-generation RNA sequencing (RNA-seq) technology is a convenient and rapid means to study gene expression at the whole-genome level and defines putative gene functions (Jain 2011; Ozsolak and Milos 2011). Several transcriptome studies have focused on the relevance of vernalization of lily bulbs to flowering (Huang et al. 2014; Liu et al. 2014; Villacorta-Martin et al. 2015). One report on the transcriptome profiles of L. longiflorum bulbs at several time points during cold exposure presented a detailed analysis of gene expression dynamics during vernalization and revealed some candidate vernalization genes in lily (Villacorta-Martin et al. 2015).

In addition to vernalization, photoperiod is also important for the flowering transition of Lilium. LD conditions could hasten the flowering of L. longiflorum. Additionally, small non-vernalized L. longiflorum bulbs could flower under LD conditions, but not under SD conditions (Lazare and Zaccai 2016). However, the molecular mechanism of the photoperiodic flowering pathway in lily has not been explored yet. And the internal transition stage from vegetative to reproductive growth, when the morphology of the SAM still remains in a vegetative status, has not been identified.

In this study, the different developmental stages of vegetative juvenile, flowering induction and floral differentiation of L. × formolongi seedlings were successfully identified under LD conditions. Additionally, plant tissues of the four stages were sampled for high-throughput Illumina sequencing to analyze the global gene expression profiles during the flowering initiation process. Based on the extensive data analyses, critical flowering homologs from several regulatory pathways were identified for the first time. The data set will serve as a foundation to understand the photoperiod-mediated, age-dependent and carbohydrate nutritional molecular mechanisms of flowering regulation, and will help to discern the molecular characteristics of the fast transition from vegetative to reproductive growth within a year from seeds of L. × formolongi.

Material and method

Plant materials and crucial stages confirmation during vegetative growth and flowering initiation

The seeds of L. × formolongi cv. Raizan 2 were sown in seedling-raising trays after stratification storage at 4 °C for a month. The seeds germinated and grew in a light growth chamber at 25/18 °C day/night with 70% humidity and were illuminated with white light (320 μmol m−2 s−1 photosynthetic photon flux density) under a 16/8-h photoperiod. At approximately 2 months after sowing, after the seedlings grew 2–3 rosette leaves, each seedling was transplanted into an 8-cm pot. At nearly 4 months after sowing, when the average number of rosette leaves was 5–6, the seedlings were transplanted into 15-cm pots. All of the seedlings were carefully monitored until flowering.

To identify the different stages of L. × formolongi seedlings during vegetative growth and flowering initiation under LD conditions, the phenotypic traits and anatomical structures of the SAM were investigated weekly from seeds germination to visible floral bud appearance. The phenotypic traits included the average number of rosette leaves before bolting and internodes after bolting, as well as the average plant height. The phenotypic traits were observed and recorded among 30 seedlings each week. Additionally, three stem apices were sampled from three seedlings with conforming phenotypic traits, and the samples were stored in Formalin–acetic acid–alcohol liquid (70% ethyl alcohol:glacial acetic acid:methanol; 18:1:1) for permanent paraffin sections to observe the course of flower differentiation at each week. Additionally, each 2 g sample of fresh leaves and stem apices were collected from the conforming seedlings and sampled for three biological replicates as the candidate materials for RNA-seq and qRT-PCR at each week. The leaves and stem apices were frozen in liquid nitrogen and stored at −80 °C separately.

Moreover, according to studies in A. thaliana, the circadian rhythm expression patterns of CO and FT are important biomarkers of flowering induction and transition (Andrés and Coupland 2012; Meng et al. 2011). The COL homolog LfCOL9 (KJ744206) and FT homolog LfFT1 (KJ744207) were cloned from L. × formolongi cv. Raizan 3 in our previous study. At the 2–3 and 9–10 rosette leaf stages and the 1–2 internode stage, the LfCOL9 and LfFT1 mRNA accumulations in the leaf blades were investigated by qRT-PCR every 4 h over 3 days. At each time point, leaves from three seedlings were collected as biological replicates and stored at −80 °C for RNA extraction.

Total RNA was extracted using the kits from Aidlab Biotechnology Co., LTD., China, according to the manufacturer’s instructions. The quality of RNA was characterized on a 1% agarose gel electrophoresis and verified using an ND-1000 Spectrophotometer (NanoDrop, Wilmington, DE). The standards applied were 1.8 ≤ OD260/280 ≤ 2.2 and OD260/230 ≥ 1.8. cDNA was synthesized using a PrimeScript Double Strand cDNA synthesis kit (Takara, Dalian, China) according to the manufacturer’s instructions. Aquaporin TIP4-1 was selected as reference gene to standardize the results because it had shown the most stable expression in different developmental stages in lily according to our previous experiment. All the primers are listed in Table S1. qRT-PCR was performed in a Bio-Rad Connect Real-Time PCR Detection System using SYBR Premix EX Taq II Kit (TaKaRa) according to the manufacturer’s protocol. All the samples were run in triplicates in 96-well optical reaction plates (Bio-rad, USA).Relative expression was determined by CT values and calculated by 2−ΔΔCT algorithm based on the expression of Aquaporin TIP4-1.

Illumina sequencing and gene functional annotation

The leaves and stem apexes from the four identified stages, including vegetative juvenile (VJ), flowering induction phase I (FI-1), flowering induction phase II (FI-2) and floral differentiation (FLD) stage were used for constructing four cDNA libraries for RNA-seq. Total RNA of the leaves and stem apexes from the same stage were extracted separately and then mixed in equal amounts, using a TRIzol reagent (Invitrogen, Carlsbad, CA, USA) according to the manufacturer’s instructions. RNA concentration and integrity number (RIN > 7.5) were evaluated on an Agilent 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA, USA). Four cDNA libraries were constructed from the samples of four stages in which equal amounts of total RNA were pooled from three biological replicates per stage. The cDNAs were synthesized according the protocols of the Beijing Genome Institute (BGI; Shenzhen, China), then the four cDNA libraries were subjected to paired-end sequencing on an Illumina HiSeq™ 2000 platform. The clean reads sequenced from every cDNA library were assembled de novo to become unigenes using Trinity platform with parameters of ‘K-mer = 25, group pairs distance = 250’ (Grabherr et al. 2011). TGICL software was used to remove redundant sequences from unigenes and assembled them as long as possible (Pertea et al. 2003). Finally, all unigenes from the four gene expression libraries were assembled again to a single set of non-redundant library called All-Unigene, which was to be a reference library for differential gene expression analysis among the four libraries.

Furthermore, All-Unigene was aligned to a series of protein databases (E-value <1e−5) including NCBI non-redundant protein database (Nr), Swiss-Prot database, Kyoto Encyclopedia of Genes and Genomes protein database (KEGG) and Clusters of orthologous Groups of proteins database (COG) by BLASTx and an NCBI nucleotide database (Nt) (E-value <1e−5) by BLASTn to obtain functional annotation for all unigenes. The best aligned results were used to decide the direction and protein coding-region (CDS) of the unigene sequences. Furthermore, all the unigenes were subjected to Blast2GO program to obtain gene ontology (GO) annotations and classifications (http://www.blast2go.org/ ).

Differential gene expression analysis, and gene ontology (GO) and Kyoto encyclopedia of genes and genomes (KEGG) pathway enrichment analyses

Clean reads from each library of the four stages were mapped to All-Unigene, which is a reference database, using the short oligonucleotide analysis package SOAP aligner/soap2. The number of fragments that were uniquely aligned was normalized using the fragments per kilobase of transcript per million mapped reads (FPKM) method to calculate the unigene expression levels (Mortazavi et al. 2008). Then, the unigene expression levels between different libraries were compared using the edgeR package (Robinson et al. 2010) for differential expression analyses in the OmicShare tools, a free online platform for data analysis (http://www.omicshare.com/tools). The differentially expressed genes (DEGs) between the different gene expression libraries of four stages were screened restrictedly by Q-value ≤0.05 and the absolute value of log2 fold chang ≥1. As there was no replicate in this study, biological coefficient of variation which was the square-root of dispersions was set to 0.01 following the suggestion of edgeR official manual. The gene expression libraries of four stages were organized into six groups, VJ vs. FI-1, FI-1 vs. FI-2, FI-2 vs. FLD, VJ vs. FI-2, VJ vs. FLD and FI-1 vs. FLD, for pairwise comparisons to identify the DEGs during different stages.

Furthermore, the DEGs between any pair of stages were extracted and clustered by Short Time-series Expression Miner software (STEM) to analyze their expression profiles during the four stages (Ernst and Bar-Joseph 2006). The gene expression data of the DEGs in the four stages were normalized to 0, log2 (FI-1/VJ), log2 (FI-2/FI-1) and log2 (FLD/FI-2). The clustered profiles with P-value ≤0.05 were considered as significantly expressed. Then, the DEGs of significantly clustered profiles were subjected to GO and KEGG pathway enrichment analyses to identify significantly enriched GO terms and metabolic pathways compared with the whole genome background. The calculated P-value underwent a false discovery rate correction, taking the corrected P-value (Q-value) ≤0.1 as a threshold. GO terms and KEGG pathways fulfilling this condition were defined as significantly enriched. Furthermore, a bubble map was used to visualize the enriched KEGG results, which showed the significantly enriched KEGG pathways, the number of enriched unigenes, the enrichment factor and the significance.

Homologs related to flowering and real-time PCR validation

This RNA-seq analysis was performed to confirm the important flowering homologs in L. × formolongi. Several approaches were followed, including an initial screening of the KEGG pathways and GO terms related to flowering. Then, the coding sequences (CDSs) of the candidate unigenes involved in these KEGG pathways and GO terms were queried against the UniProtKB database using the BLAST algorithm (Villacorta-Martin et al. 2015). The unigenes with top hits were then identified as putative homologs based on the highest percentage of query coverage (≥70%) and the protein identity (≥60%), as well as the BLAST e-value (≤0.00001). Another approach, based on the functional annotations of all of the unigenes, selected those with annotation terms involved in flowering. Then, their CDSs were queried against the UniProtKB database using the BLAST algorithm, and the unigenes with top hits were selected and filtered as described previously.

Furthermore, the flowering homologs that showed differential expression were defined by a rigorous algorithm with an Q-value ≤0.05 and an absolute value of log2 fold change ≥1 as thresholds. Additionally, Cytoscape 3.2 software was used to generate and visualize co-expression clusters of the DEGs related to flowering using a correlation coefficient value of 0.8. Furthermore, protein sequences of important flowering homologs from L. × formolongi, A. thaliana and Oryza sativa were aligned using ClustalW in MEGA6 with default parameters. The phylogenic tree of MADS-box family genes was generated using the Bayesian method by MrBayes-3.2.6. Phylogenic analyses of other flowering homologous families used the Maximum Likelihood or Neighbor-Joining methods in MEGA6. The bootstrap analysis employed 1000 replicates, and the Poisson model was used for substitutions.

To validate the reliability of the RNA-seq analysis and the expression profiles of the important homologs related to flowering, 12 homologs were selected for qRT-PCR. The plant tissues of three individuals from the four stages were sampled at the same time with those used for RNA-seq.

Furthermore, in order to evaluate the biological variation of the gene expression levels in new material, three important homologs were chosen to evaluate their gene expression levels in three individuals as biological replicates different from the material used for RNA-seq. A new batch of seeds were sown, and the plant tissues of three biological replicates were sampled from eight development stages from seeds germination to the appearance of floral buds under LD conditions. The total RNA of the leaves and stem apices from each individual were extracted separately and then mixed in equal amounts to synthesize the cDNAs for qRT-PCR tests. The RNA extraction and cDNA synthesis methods were described previously, as were the preparation and amplification program for qRT-PRC. Aquaporin TIP4-1 was used as the internal control to standardize the results. All of the primers of the validated genes are listed in Table S1.

Results

Confirmation of the crucial stages during vegetative growth and flowering initiation

The phenotypic traits, anatomical SAM structures and the circadian rhythm expression patterns of LfCOL9 and LfFT1 in L. × formolongi cv. Raizan 2 seedlings were investigated from seed germination to visible floral bud appearance to define the crucial stages during vegetative growth and flowering initiation under LDs (Fig. 1a–c).

Fig. 1
figure 1

Confirmation of different developmental stages during vegetative growth and flowering initiation process. a The morphological dynamic profile of L. × formolongi during different stages of vegetative growth and flower development. Bar 30 mm. b The anatomic structures of SAM using paraffin section during 2–3 rosette leaves (VJ), 9–10 rosette leaves, 1–2 internodes (FI-1), 5–6 internodes (FI-2) and 9–10 internodes (FLD). Bar 0.5 mm. c The diurnal circadian rhythm expression patterns of LfCOL9 and LfFT1 in leaves under LDs during the stages of 2–3 rosette leaves (VJ), 9–10 rosette leaves and 1–2 internodes (FI-1). Relative expression levels were determined by qRT-PCR (y-axis). The x-axis represents time point (hours). Data points represent an average of three biological replicates with three technical replicates. Error bars represent SE. The shaded bars over each chart represent dark periods

Full size image

In seedlings with 2–3 to 9–10 rosette leaves, the morphology of the SAM exhibited a hemispheric vegetative shape (Fig. 1b). To ensure the accuracy of the VJ stage used for RNA-seq, the 2–3 rosette leaf stage was selected for a gene expression analysis. The mRNA accumulations of LfCOL9 and LfFT1 did not show diurnal rhythmic expression patterns for 3 days at that stage (Fig. 1c). Thus, the 2–3 rosette leaf stage was selected as the representative of the VJ stage for RNA-seq.

As the seedlings grew to have 9–10 rosette leaves at the 23rd week, the morphology of the SAM remained vegetative (Fig. 1b), while the expressions of LfCOL9 and LfFT1 started to display ambiguous circadian rhythms and low amplitudes (Fig. 1c). When the seedlings underwent bolting at the 24th week, they possessed 1–2 internodes and the morphology of the SAM was still hemispheric vegetative (Fig. 1b). However, the expressions of LfCOL9 and LfFT1 showed distinct and strong circadian rhythm patterns for 3 days with transcript levels peaking 4 h after dark. In addition, the diurnal expression of LfCOL9 also had a transcript peak 12 h after illumination (in the light) (Fig. 1c). Meanwhile, the transcript accumulation of LfFT1 was especially up-regulated compared with at the 9–10 rosette leaves stage (Fig. 1c). Therefore, the period from 9 to 10 rosette leaves to 1–2 internodes (just before and after bolting) was identified as flowering induction phase I (FI-1), and the tissues of the 1–2 internode stage (just bolting) was selected to represent FI-1 for RNA-seq.

When the seedlings had 5–6 internodes during the 25th week, paraffin sections showed that the morphology of the SAM appeared to be elongated and broad (Fig. 1b). This indicated that the SAM had undergone the morphological transition from vegetative to flowering, which was identified as flowering induction phase II (FI-2), and plant tissues of this stage were used for RNA-seq.

When the internode numbers reached 9–10 during the 26th week, the SAM enlarged and developed into inflorescence primordial (Fig. 1b). When the seedlings grew to 16–17 internodes, the SAM became concave and developed into small floral primordial (Fig. 1b). Thus, at the 9–10 internode stage, the seedlings entered into the FLD stage, the plant tissues of this stage were also used for RNA-seq.

Differential gene expression dynamics, GO and KEGG pathway enrichment analysis

To profile the global gene expression profile during the flowering initiation process of L. × formolongi, the four cDNA libraries of tissues from the VJ, FI-1, FI-2 and FLD stages were separately sequenced on the Illumina Hiseq 2000 platform. A total of 52,824 unigenes were generated from the four libraries, of which 37,032 were differentially expressed. The DEGs contained two expression patterns, up-regulated and down-regulated. Most of the DEGs were up-regulated in the FI-1 (14,713) and FLD (8310) stages, but down-regulated in the FI-2 stage (10,792) compared with the VJ stage (Fig. 2a).

Fig. 2
figure 2

Differential gene expression patterns and the KEGG pathway enrichment analysis results. a The number of the DEGs obtained from the pairwise comparison of four stages. The red columns represent the numbers of up-regulated DEGs in each pairwise comparison, and blue columns represent the down-regulated numbers. b Cluster analysis of the expression profiles of the DEGs. The profile ID and P-value are on the upper side of each profile block. The number in the bottom left indicates the number of genes assigned to the profile during clustering. Profiles blocks with colored background are significant in terms of the P-value. c A bubble map shows the significantly enriched results of KEGG pathways among the DEGs of significantly clustered profiles. The y-axis shows the top 15 enriched KEGG pathways. And the 15 pathways were sorted by the significance level from the bottom to the top on the y-axis. The x-axis shows the enrichment factor (the enriched gene number proportion of background gene number) of each enriched pathway. The color of the round dots represents the Q-value of each pathway and the size represents the number of enriched unigenes

Full size image

To analyze the stage-specific expression profiles of the identified DEGs, the 37,032 DEGs were clustered by STEM according to their expression profiles. The result indicated that all of the DEGs could be clustered into 10 profiles, and just 4 profiles (colored blocks) contained 29,910 DEGs having significant expression patterns (P-value ≤0.05) (Fig. 2b). And profiles 1, 4 and 7 were the most significantly clustered.

The 29,910 DEGs of four significantly clustered profiles were subjected to GO and KEGG pathway enrichment analysis. The results of all the enriched GO terms are listed in Tables S2–S4. A bubble map indicates that 15 pathways were significantly enriched and the most enriched KEGG pathway was phenylalanine metabolism (Fig. 2c).

In addition, the DEGs of the significantly clustered profile 7 were subjected to a GO enrichment analysis. Some significantly enriched GO terms in the biological process category were related to reproductive transition and flower development (Table 1), of which reproduction (GO:0000003), DNA methylation (GO:0006306), floral organ morphogenesis (GO:0048444), flower development (GO:0009908) and reproductive shoot system development (GO:0090567) were the most significantly enriched GO terms. The results suggest that most of the DEGs related to flowering showed the expression pattern of profile 7, namely up-regulated in the FI-1 stage, then down-regulated in the FI-2 stage, and finally up-regulated again in the FLD stage. Thus, we hypothesized that FI-1 may be the key switching stage in which many important genes that respond during flowering were induced and up-regulated. The genes that have their highest transcript accumulations in the FLD stage may be related to floral meristem development.

Table 1 The significantly enriched GO terms related to reproductive transition and flower development in biological process of profile 7
Full size table

Differential expression of putative homologs related to flowering in L. × formolongi

In total, 122 homologs implicated in flowering were identified from the database, and 85 homologs showed significantly differential expression during the four stages. These identified DEGs could be distributed into five genetic flowering pathways, photoperiod, age-dependent, trehalose-6-phosphate (T6P), gibberellin and ambient temperature (Table 2). The DEGs of the photoperiod pathway (41) were the most represented, followed by the age-dependent pathway (13) and the T6P pathway (9). Only one putative homolog each was identified in the gibberellin and the ambient pathways. Moreover, another 20 homologs were implicated in floral meristem identity. Table 2 shows the numbers and descriptions of important DEGs related to flowering in L. × formolongi, as well as the accession number and expression profile of the gene with the highest homology. The other putative flowering homologs are described in Table S5.

Table 2 The putative homologs related to flowering and their expression profiles in mixed sample of leaves and stem apexes during four stages in L. × formolongi. Gene expression during different stage is schematically represented based on the log-fold change from VJ to FLD stage
Full size table

A co-expression network was used to depict the differential expression profiles of all of the flowering homologs during flowering initiation (Fig. 3). The homologs could be clustered into three expression types in which closely related genes displayed conserved expression profiles. The predominant type I cluster included genes that showed the highest expression in the FI-1 stage, like the circadian clock homologs CRY1 and PHYTOCHROME A (PHYA), several COL and FT homologs from the photoperiod pathway, as well as several TPS homologs from the T6P pathway (Fig. 3). Cluster type II included the genes that exhibited the highest transcription levels in the VJ stage, such as the rest of the circadian clock and COL homologs from the photoperiod pathway, and the other TPS family members from the T6P pathway (Fig. 3). Cluster type III consisted of genes up-regulated in the FLD stage, like the floral meristem identity homolog APETALA1, LEAFY, and several AGAMOUS-LIKE (AGL) and SPL homologs (Fig. 3).

Fig. 3
figure 3

The co-expression network map depicting the differential expression profiles of the flowering homologs and TF genes during flowering initiation. Cluster type I represents the up-regulated homologs in the FI-1 stage, cluster type II represents the homologs showed the highest expression in the VJ stage and cluster type III represents the up-regulated homologs in the FLD stage. The orange circles represent the members of MADS-box TF family, yellow circles represent the genes of SBP TF family, and the blue circles represent the genes of CO-like TF family. Remaining pink and blank circles represent the other TF family members and general homologs respectively

Full size image

Differential expression of the TF genes related to flowering

TFs are the key proteins that mediate transcriptional regulation and also play crucial roles as molecular switches for gene expression in flowering transition. According to the previous study in A. thaliana, most of the important regulatory genes implicated in flowering transition encode TFs (Par̆enicová et al. 2003; Singh et al. 2013). Therefore, we searched for TF genes related to flowering in L. × formolongi and studied their expression dynamics. Among the identified putative homologs, 52 genes encoded for TFs and showed differential expression, including the members of the bHLH, bZIP, RAV, MYB, Dof, CO-like, AP2/EREBP, SBP, GRAS, FLO_LFY and MADS-box TF families. The co-expression network also depicted the differential expression profiles of the TF genes during the four stages (Fig. 3).

In the network, the MADS-box TF family genes (orange circles) were the most represented among the various families, with 20 genes showing differential expression. They were mainly distributed in clusters I and III, representing up-regulated profiles in the FI-1 and FLD stages, respectively (Fig. 3). The SBP TF family genes (SPLs, yellow circles) (11) were the second most represented and were also mainly distributed in clusters I and III (Fig. 3). They were followed by CO-like TF family members (blue circles, eight), which were distributed in clusters I and II (Fig. 3), and exhibited their highest transcription levels in the FI-1 and VJ stages. The other TF family genes (pink circles) were mainly distributed in cluster I and were namely up-regulated in the FI-1 stage (Fig. 3).

These findings suggested that the FI-1 stage was the key stage of flowering transition and that most TF genes were up-regulated in this stage. Remarkably, the up-regulated MADS-box, SBP and CO-like TF genes in the FI-1 stage may be the critical trigger factors for flowering induction and transition.

In addition, we analyzed the differential expression profiles of the identified MADS-box homologs, including SUPPRESSOR OF OVEREXPRESSION OF CO 1 (SOC1), AGL, AP1, SHORT VEGETATIVE STAGE (SVP) (Fig. 4b). An SVP homolog and four SOC1 homologs that showed the highest transcription levels in the FI-1 stage may be involved in the flowering transition. Two AGL homologs, LfAGL44 and LfAGL19, and another three SOC1 homologs, LfSOC1.1, LfSOC1.2 and LfSOC1.3, were not only up-regulated in the FI-1 stage, but also showed high expression levels in the FLD stage, which means these genes may be related to the flowering transition process, but also participated in the floral development. Furthermore, six AGL homologs, LfAGL1, LfAGL2, LfAGL3, LfAGL4, LfAGL24, LfAGL15 and three AP1 homologs, LfAP1.1, LfAP1.2 and LfAP1.3, were all up-regulated in the FLD stage and may function as floral meristem identity genes (Fig. 4b).

Fig. 4
figure 4

The phylogenetic relationship and differential expression profiles of the putative MADS-box homologs. a Bayesian tree of the identified MADS-box TF genes of L. × formolongi and A. thaliana. The gene names colored in pink were A. thaliana MADS-box TF genes and the black were homologs from L. × formolongi. The prefixed green dots represent the highest expression in the VJ stage, red dots represent the up-regulated expression in the FI-1 stage, and blue dots represent up-regulated expression in the FLD stage. b A heatmap exhibiting the differential expression profiles of all the MADS-box homologs. Gene names are colored according to their expression patterns. Green color represents the highest expression in the VJ stage, red represents the up-regulated expression in the FI-1 stage, and blue represents the up-regulated expression in the FLD stage

Full size image

Differential expression of flowering homologs in the photoperiod pathway

A total of 41 DEGs related to circadian rhythm and photoperiod-mediate flowering were identified, including the circadian rhythm genes CRY1, PHYA, TIMING OF CAB EXPRESSION 1 (TOC1), FLAVIN-BINDING KELCH REPEAT F-BOX 1 (FKF1), the rhythm output genes GI, CO and FT, and the downstream bZIP transcription factor FD (Table 2). A heatmap showed the differential expression profiles of all 41 homologs (Fig. 5c). Almost all of the circadian clock homologs exhibited high transcript accumulations in the VJ or FI-1 stages, including LfGI, the homolog of GI, which had its highest expression in the VJ stage but also maintained a high expression level in the FI-1 stage (Fig. 5c).

Fig. 5
figure 5

The phylogenetic relationship and differential expression profiles of the putative photoperiod homologs. a, b Maximum likelihood trees of the identified COL and FT-like homologs of L. × formolongi, A. thaliana and O. sativa. Bootstrap values from 1000 replicates are used to assess the robustness of the tree. The scale indicates the average number of substitutions per site. The gene names colored in pink are A. thaliana genes, blue are O. sativa genes and the black are the homologs from L. × formolongi. The prefixed green dots represent the highest expression in the VJ stage, red dots represent the up-regulated expression in the FI-1 stage, and blue dots represent up-regulated expression in the FLD stage. The structure diagrams in the right side show the domain structures of B-box motif (red rectangles), second divergent B-box motif (gray rectangles), CCT motif (yellow pentagons) of the COL genes from L. × formolongi, A. thaliana and O. sativa. c A heatmap exhibiting the differential expression profiles of all the photoperiodic homologs. Gene names of LfCOLs and LfFTs are colored according to their expression patterns during the four stages refer to the description in Fig. 4

Full size image

In A. thaliana, CO is a member of COL gene family that contains 16 other members (Griffiths et al. 2003). In this study, eight COL homologs members were identified. Based on the phylogenetic relationships and conserved protein domains of the COL family members between L. × formolongi, A. thaliana and O. sativa, the eight COL homologs were divided into three types (Fig. 5a). The first type contained two B-box motifs and only included LfCOL5. The second type exhibited one B-box motif, and consisted of LfCOL6, LfCOL7, LfCOL13, LfCOL14, LfCOL15 and LfCOL16. The last type contained a normal B-box motif and a second divergent B-box motif, and only included LfCOL9 (Fig. 5a). Thus, the LfCOLs homologs with one B-box motif were predominant. Additionally, the LfCOLs were phylogenetically close related to the O. sativa COLs, most of which also contained only one B-box motif (Fig. 5a).

Moreover, among the eight LfCOLs, five homologs, including LfCOL7 and LfCOL9, had their highest expression levels during the VJ stage (Fig. 5c). The other three homologs, LfCOL5, LfCOL6 and LfCOL14, were up-regulated in the FI-1 stage (Fig. 5c). In addition, three CYCLIC DOF FACTOR 3 (CDF3) homologs, a CONSTITUTIVE PHOTOMORPHOGENIC 1 (COP1) homolog and two SUPPRESSOR OF PHYA-105 1 (SPA1) homologs, which as the repressors of COL or the COL protein, exhibited high expression levels in the FI-1 stage (Fig. 5c).

FT is a key flowering activator, and protein TWIN SISTER of FT-like (TSF) is the most homologous member of the same gene family as FT in A. thaliana. Moreover, Heading date 3a (Hd3a) is the rice (O. sativa) homolog of the A. thaliana FT (Kojima et al. 2002). Two L. × formolongi putative homologs of FT-like were identified in this study. The phylogenetic relationship of the FT and TSF homologs among L. × formolongi, A. thaliana and O. sativa were analyzed using a maximum likelihood tree (Fig. 5b). The results indicated that LfFT1 was most closely related to Hd3a and that LfFT2 was closely related to OsTSF (Fig. 5b). LfFT1was up-regulated in the FI-1 stage, and LfFT2 was highly expressed in both the FI-1 and FLD stages (Fig. 5c).

In addition, two TEMPRANILLO 1 (TEM1) homologs, LfTEM1.1 and LfTEM1.2, which are repressors of FT and encode members of the RAV transcription factor family in A. thaliana, were up-regulated in the FI-1 stage (Fig. 5c). The FD homolog, LfFD, showed the highest expression in the FI-1 stage (Fig. 5c). FD is an important activator in the triggering of flowering at the stem apex (Wellmer et al. 2006).

Differential expression of the age-dependent and T6P pathway homologs

A set of 13 DEGs belonging to the age-dependent pathway were identified, including 11 SBP TF family member SPLs, whose homologs in A. thaliana act as the target genes of miR156, and 2 AP2/EREBP TF family member AP2s, whose homologs act as the target genes of miR172. In addition, nine TPS homologs belonging to the T6P signal pathway were identified.

Among the 11 SPL homologs, seven homologs were up-regulated in the FI-1 stage, such as LfSPL3, three homologs were up-regulated in the FLD stage, such as LfSPL9. Only LfSPL13 not only up-regulated in the FI-2 stage, but also showed a high expression level in the FLD stage (Fig. 6c). Additionally, among the seven up-regulated LfSPLs in the FI-1 stage, five, such as LfSPL16 and LfSPL17, also showed high expressions in the FLD stage (Fig. 6c). The two AP2 homologs were all up-regulated in the FI-1 stage (Fig. 6c).

Fig. 6
figure 6

The phylogenetic relationship and differential expression profiles of the putative SPL, AP2 and TPS homologs. a Maximum likelihood tree of the identified TPS homologs between L. × formolongi, A. thaliana and O. sativa. The color of the gene names and the prefixed dots refer to the description in the previous Fig. 5. b Neighbor-joining tree of the identified SPL homologs between L. × formolongi and A. thaliana. The gene names colored in pink are A. thaliana genes, and the black are the homologs from L. × formolongi. The prefixed red dots represent the up-regulated expression in the FI-1 stage, yellow dot represents up-regulated expression in the FI-2 stage and blue dots represent up-regulated expression in the FLD stage. c A heatmap exhibiting the differential expression profiles of the identified LfTPS, LfSPL and LfAP2 homologs. Gene names are colored according to their expression patterns. Green color represents the highest expression in the VJ stage, red represents the up-regulated expression in the FI-1 stage, purple represents the up-regulated expression in the FI-2 stage and blue represents the up-regulated expression in the FLD stage

Full size image

Of the nine TPSs homologs, LfTPS1 was clustered with AtTPS1 and OsTPS1 in the phylogenetic tree and exhibited the highest transcription level in the VJ stage (Fig. 6a, c). The other LfTPSs either showed their highest expression levels in the VJ stage, or were up-regulated in the FI-2 stage (Fig. 6c). However, only LfTPS9 showed its highest expression level in the FLD stage (Fig. 6c).

Verification of the gene expression profiles by qRT-PCR

To validate the accuracy and reproducibility of the expression profiles of genes in the RNA-seq analysis, 12 important DEGs related to flowering transition and floral meristem identity were selected for qRT-PCR. The candidate DEGs included LfCOLs, LfFT1, LfSPLs and LfTPSs, which acted as the key regulators in the photoperiod, age and T6P pathways. The linear regression analysis of the fold change of the gene expression between qRT-PCR and RNA-seq showed significantly positive correlation (correlation coefficient of 0.72, p-value was 1.16E−10) (Fig. 7a). And all of the homologs showed expression patterns in qRT-PCR analysis that were consistent with the RNA-seq data, confirming the reliability of our RNA-seq analysis (Fig. 7b).

Fig. 7
figure 7

Coefficient analysis of fold change data between qRT-PCR and RNA-seq. a Correlation of gene expression results obtained from qRT-PCR and RNA-seq of 12 selected genes during four stages. Scatterplots were generated by the log2 expression ratios from RNA-seq (y-axis) and qRT-PCR (x-axis). b Expression profiles of 12 putative flowering homologs revealed by qRT-PCR (left y-axis) and RNA-seq (right y-axis ) during the four stages. Data from qRT-PCR were means of three biological replicates with three technical replicates and bars represent SE. c Expression profiles of three putative flowering homologs in three individuals revealed by qRT-PCR (left y-axis) during eight stages. The notes of x-axis represent various stages VJ 2–3 rosette leaves; 1 5–6 rosette leaves; 2 9–10 rosette leaves; FI-1 1–2 internodes; FI-2 5–6 internodes; 3 7–8 internodes; FLD 9–10 internodes; 4 20–21 internodes. Data from qRT-PCR were means of three technical replicates and bars represent SE

Full size image

Furthermore, three important flowering homologs of LfCOL5, LfCOL9 and LfFT1 were chosen to evaluate their expression levels in three individuals during eight developmental stages in of L. × formolongi under LD conditions. The materials were different from the RNA-seq, and sampled from a new batch of seedlings from eight development stages, including 2–3 rosette leaves (VJ), 5–6 rosette leaves, 9–10 rosette leaves, 1–2 internodes (FI-1), 5–6 internodes (FI-2), 7–8 internodes, 9–10 internodes (FLD) and 20–21 internodes (Fig. 7c). The three replicates of each homolog all showed consistent expression patterns during the eight stages. The three genes expression patterns of the three individuals in the VJ, FI-1, FI-2 and FLD stages were consistent with the RNA-seq data of the same stages (Fig. 7c). Thus, the biological variation of the gene expression levels in the RNA-seq is reliable to some extent.

Discussion

Confirmation of different stages during flowering initiation

Lilium × formolongi is hallmarked of its ability to flower within a year after sowing. The confirmation of the flowering induction stage is very important for understanding the genetic mechanism of the prompt transition from vegetative growth to flowering in L. × formolongi. However, the flowering induction stage is hard to identify because the SAM is still morphologically vegetative, while the internal physiological mechanism and molecular dynamics have changed dramatically and switched to the reproductive stage (Meyerowitz et al. 1991). According to studies in the typical LD plant A. thaliana, the circadian rhythm expression patterns of CO and FT are crucial in triggering the flowering induction and transition (Andrés and Coupland 2012; Suárez-López et al. 2001). Our study suggests that the circadian rhythm expression patterns of LfCOL9 and LfFT1 in L. × formolongi could be applied as biomarkers of the flowering induction stage under LD conditions.

The circadian rhythm expression patterns of LfCOL9 in the FI-1 stage almost conformed to the diurnal expression pattern of CO in A. thaliana (Andrés and Coupland 2012). Nevertheless, the expression pattern of LfFT1, with a transcript abundance peak at night, was not in agreement with the discovery in A. thaliana that showed the transcript abundance peak of FT occurred at dusk (Andrés and Coupland 2012). This may be due to the different origins, evolutionary histories and habitats of the different species. The analyses of CO and FT expressions in aspen trees had already detected variations in their relative peaks during the day–night cycle, such that southern trees showed an earlier peak than northern trees (Böhlenius et al. 2006).

In the flowering time control study of A. thaliana, the timing of flowering was carefully measured by counting the number of leaves formed before the initiation of flowering (Andrés and Coupland 2012). In this study, based on the investigation of the internal gene expression patterns and external anatomical SAM structures during different growth stages of the L. × formolongi seedlings, the corresponding phenotypic traits, including the average number of rosette leaves and internodes, could be used as the favorable morphological markers to judge the VJ, FI and FLD stages. The rosette stage, from 2–3 to 7–8 leaves, represented the vegetative growth stage of the seedlings. The stage just before and after seedling bolting was identified as FI-1. When the seedlings grew up to 5–6 internodes, the stage was identified as the morphological transition stage from vegetative to reproductive growth, namely FI-2. The FLD was identified using an average internode number of 9–10. Confirmation of the convenient morphological markers of the different growth and developmental stages would be helpful for the cultivation and flowering regulation analysis of L. × formolongi.

The differential gene expression profile in the FI-2 stage

In our RNA-seq study, 37,032 DEGs were generated from the four developmental stages of L. × formolongi seedlings under LD conditions. The global expression profiles of most DEGs were most predominant in the FI-1 and FLD stages, which demonstrated that FI-1 was the key stage involved in flowering induction and that the FLD stage involved the development of the floral meristem.

Nevertheless, the RNA-seq analysis revealed that most DEGs were down-regulated in the FI-2 stage in which the SAM morphology reached the early reproductive transition status. In A. thaliana, a global gene expression analysis during early flower development revealed that an important step for the onset of flower formation was characterized by a massive down-regulation of genes in incipient floral primordia, followed by a predominance of gene activation during the differentiation of floral organs (Wellmer et al. 2006). The results were in agreement with our transcriptome’s DEG dynamics during the flowering initiation process. The DEGs were down-regulated in the FI-2 stage and exhibited higher transcription again in the FLD stage.

In addition, some important flowering activators identified in our study showed low expression levels in the FI-2 stage, and even in the FLD stage. Recent studies in A. thaliana showed that AP1 acts predominantly as a transcriptional repressor during the initial stage of flower development. In particular, AP1 down-regulates genes that participate in flowering transition, including some of its own activators, such as SOC1, AGL24, SPL9, LFY and FD (Wellmer and Riechmann 2010). This may also explain why LfLFY and several LfSOC1s genes showed lower transcript accumulations in the FI-2 and FLD stages. Thus, FI-2 is the important transition stage for floral development, even though most DEGs were down-regulated in this stage. Moreover, some flowering activators may be repressed by LfAP1s, therefore, showing low expression levels in the FI-2 and FLD stages.

Crucial flowering pathways and regulatory genes identified in L. × formolongi

In our RNA-seq study of the flowering initiation process of L. × formolongi seedlings, 85 DEGs related to flowering were identified. This was the first discovery of most of them in lily, and they were distributed among the photoperiod, age-dependent and T6P pathways. Remarkably, some homologs may be the key activators and TF genes reported by preceding studies in A. thaliana, such as members of the LfCOL and LfFT families from photoperiod, and the LfSPL and LfTPS families from the age-dependent and T6P pathways, respectively.

In A. thaliana, the members of the COL family contain one or two zinc finger B-boxes at the amino terminus and a CCT domain at the carboxy terminus, which could be divided into three types by the conservation of the B-box domains (Robson et al. 2001). The first type has two B-box motifs and includes CO and COL1–COL5. The second type contains only one B-box motif, as in COL6–COL8 and COL16. The last type shows one B-box motif and a second divergent B-box motif, and consists of COL9–COL15 (Griffiths et al. 2003). Recent studies showed that not all of the COL gene members could regulate the flowering time. COL1 and COL2 were predicted to encode zinc finger proteins with almost 67% amino acid identities to the protein encoded by CO. However, unlike CO, the altered expression of COL1 or COL2 in transgenic plants has little effect on flowering time (Ledger et al. 2001). Moreover, COL3 is a positive regulator of red light signaling and root growth, and the over-expression of COL5 can induce flowering in SD grown A. thaliana (Datta et al. 2006; Hassidim et al. 2009).

Based on the rice genomic sequence, 16 OsCOL genes have been identified (Griffiths et al. 2003). However, we only discovered 13 OsCOLs in the NCBI and plant transcription factor database (http://planttfdb.cbi.pku.edu.cn/family.php?sp=Osj&fam=CO-like) that showed integrated protein domains. The functions of these OsCOLs were different. Hd1 is the homolog of A. thaliana CO and a major determinant of photoperiod sensitivity in O. sativa. A report indicated that OsCOL3 negatively controlled flowering time under SDs by repressing the expression of Hd3a, independent of the SD-promotion pathway (Kim et al. 2008).

Lily, like O. sativa, is a monocotyledon. In our study, the eight identified LfCOL homologs could also be divided into three types by the conservation of the B-box domains, as in A. thaliana. However, most of the LfCOLs were classified into the second type that contained only one B-box motif, similar to O. sativa. Furthermore, the expression profiles of the eight LfCOLs were different. Several homologs showed their highest expression levels in the VJ stage, while others were up-regulated in the FI-1 stage or showed high expression levels in both the VJ and FI-1 stages. LfCOL9, which had been cloned in our previous study from L. × formolongi and showed a circadian rhythm expression pattern in the FI-1 stage, exhibited high expression levels in both the VJ and FI-1 stages. Thus, LfCOLs may great influence the induction of flowering, but the functions of various LfCOL family members may be different and complex. Determining which putative LfCOL homolog is the most homologous to A. thaliana CO or O. sativa Hd1, and as the critical flowering activators that promote flowering under LDs, will require more intensive studies in the future.

FT is a pivotal integrator in A. thaliana because almost all of the flowering regulatory pathways converge on this gene, and FT transmits the floral inductive signal to downstream floral identity genes, including LFY and AP1 (Mouradov et al. 2002). The FT/TERMINAL FLOWER 1 (TFL1) family is a small gene family whose members, including FT, TFL1, TSF, protein CENTRORADIALIS-like (ATC), MOTHER OF FT AND TFL1 (MFT) and BROTHER OF FT AND TFL1 (BFT) (Yoo et al. 2010). Among these family members, TSF is most homologous to FT, and it shows a similar diurnal oscillatory expression pattern and produces an early-flowering phenotype when overexpressed (Michaels et al. 2005; Yamaguchi et al. 2005).

Lilium. × formolongi is a LD monocot plant, and LfFT1 and LfFT2 were both up-regulated in the FI-1 stage. LfFT1 was phylogenetically closely related to O. sativa Hd3a, while LfFT2 was closely related to O. sativa TSF. Thus, determining which LfFT homolog predominantly functions like the O. sativa Hd3a or A. thaliana FT to trigger flowering under LDs will require more studies in the future.

Putative gene networks controlling flowering induction and development in L. × formolongi

Studies on the genetic mechanism of flowering time in the model plant A. thaliana have shown that the responses to various external and internal conditions are integrated by a complex gene regulatory network, which includes several regulatory pathways that control flowering transition (Wellmer and Riechmann 2010). The genetic basis of flowering induction and floral development in the model plant A. thaliana appears to be similar to those in other plants, but species differences among the different regulatory pathways still exist (Khan et al. 2014). A hypothetical schematic diagram was created to depict the genetic and molecular interactions involved in photoperiod, age and T6P responses in L. × formolongi based on the annotation of the flowering homologs and their stage-specific expression profiles obtained from the RNA-seq data (Fig. 8a, b).

Fig. 8
figure 8

Putative gene regulatory network of flowering induction and development, as well as the differential expression profiles of most flowering homologs of L. × formolongi. a The putative network of photoperiodic, age-dependent and T6P flowering pathways in L. × formolongi. Gene names are colored according to their expression patterns. Green color represents the highest expression in the VJ stage, red represents the up-regulated expression in the FI-1 stage, and blue represents the up-regulated expression in the FLD stage. And gene names are followed by the numbers of differentially expressed genes. Arrows in red dotted lines represent gene activation and black blunted dotted lines represent gene repression. b A heatmap exhibiting the differential expression profiles of most identified homologs in photoperiod, age-dependent and T6P pathways. Gene names are colored according to their expression patterns refer to the previous description in Fig. 6

Full size image

Arabidopsis thaliana is a facultative LD plant, in which inductive photoperiods are perceived in the leaves and the circadian clock acts as a timekeeping to measure the day length (Samach and Coupland 2000). The simplest circadian clock models consist of photoreceptors, core oscillators and rhythm output pathways (Gardner et al. 2006). GI, CO and FT are distinct regulatory genes of the circadian clock output pathway, and GI promotes the expression of CO and FT under LD conditions (Mizoguchi et al. 2005). The expression of FT in the leaf vasculature is a main determinant of the timing of flowering. The FT protein is transported through the vasculature to the apex (Turck et al. 2008), where it interacts with the bZIP transcription factor FD and, coordinately, they up-regulate the MADS-box genes SOC1, AP1 and LFY (Wellmer and Riechmann 2010). The miR156-SPL/age pathway acts both in the leaves and SAM. SPL proteins promote FT expression in leaves by regulating the expression of some MADS-box TFs, such as SOC1 (Wahl et al. 2013), while at the shoot apex, SPLs and FT/FD converge on an overlapping set of targets, where SPLs directly activate the floral meristem identity genes AP1 and LFY (Wang et al. 2009; Yamaguchi et al. 2009). The AP2 domain-containing protein genes normally act as floral repressors that down-regulate the expression of SOC1 and AP1 (Aukerman and Sakai 2003). Additionally, the T6P pathway regulates flowering at two sites in the plants. In the leaves, TPS1 promotes flowering by activating the expression of FT. Furthermore, the T6P pathway controls the expression of several SPLs in the SAM (Wahl et al. 2013).

Thus, the photoperiod, age-dependent and T6P pathways are inter-connected both in the leaves and SAM, and converge through the key integrator FT gene, promoting the same targets, the floral meristem identity genes SOC1, LFY and AP1, to trigger flowering in the SAM.

In comparison with the genetic basis of flowering regulation in A. thaliana, the molecular regulatory network of flowering in L. × formolongi exhibited particular temporal and spatial expression patterns. The photoperiod pathway’s circadian clock homologs may be the first gene set to perceive the light signals and circadian rhythms in the leaves and promote the expression of LfGI and LfCOLs in the VJ and FI-1 stages, which may be related to the prompt transition of flowering in L. × formolongi. Then, LfCOLs activate the high expression levels of LfFT1 and LfFT2. The proteins of LfFT1 and LfFT2 were transported into the SAM and interacted with LfFD to activate LfLFY and some LfSOC1s in the FI-2 stage. Meanwhile, the LfTPS family members from the T6P pathway may also promote the expression of FT in the leaves, and further activate the expression of some LfSPL members in the SAM during the FI-1 stage. The up-regulated LfSPL members in the FI-1 stage may promote the expression of LfLFY, while the other LfSPLs that are up-regulated in the FLD stage may interact with the floral meristem identity LfAP1 genes to promote flowering in the SAM. While the identified LfAP2s and LfSVP homologs may act as the repressors of the floral meristem identity LfAP1 and LfSOC1 genes (Fig. 8).

Conclusion

To our knowledge, this study is the first report showing genome-wide transcriptome dynamics during the photoperiod-mediated flowering initiation process in lily. In this study, we presented a detailed analysis of gene expression dynamics in L. × formolongi, covering the stages of vegetative and reproductive development, and the transition stages identified from molecular dynamics and morphological changes. This analysis revealed a large set of putative flowering genes, which were mostly discovered in lily for the first time and may play critical roles in the flowering transition of L. × formolongi. The detailed investigations of the candidate genes in the critical photoperiod, age-dependent and carbohydrate nutritional T6P pathways in this study will be very helpful in understanding the molecular mechanisms involved in flowering. Furthermore, the candidate homologs will provide references for the breeding of lilies with short vegetative periods and for reverse genetics approaches.