Insights into the Major Metabolites Involved in the Underground Floral Differentiation of Erythronium japonicum

Background Erythronium japonicum Decne (Liliaceae) is an early spring ephemeral with an underground initial floral differentiation stage. The flowering mechanism is crucial in ornamental plants due to the associated economic value. Therefore, this study is aimed at exploring the metabolic landscape during floral differentiation, including flower primordium, perianth, stamen, and the pistil differentiation period, in E. japonicum coupled with a conjoint analysis of the metabolome and transcriptome. Using ultraperformance liquid chromatography-tandem mass spectrometry, we identified 586 metabolites from 13 major metabolite classes. Comparative metabolomics between different floral developmental stages revealed several abundant metabolites during the respective phases. Upaccumulation of p-coumaroylputrescine, scopoletin, isorhoifolin, cosmosiin, genistin, and LysoPC 15 : 0 emphasized the significance of these compounds during flower development. Furthermore, previously identified DEGs, viz., EARLY FLOWERING 3, Flowering locus K, PHD finger-containing protein, and zinc finger SWIM domain-containing protein for floral differentiation, depicted a high correlation with lipid, flavonoid, and phenolics accumulation during floral developmental stages. Conclusions Together, the results improve our interpretation of the underground floral development in E. japonicum.


Introduction
Erythronium japonicum, native to Asia, is an early spring ephemeral recently domesticated for its commercial use [1,2]. Northeast China, Japan, and Korea are considered the geographic origin of E. japonicum [1,3]. The reproductive phase in E. japonicum starts underground without vernalization and photoperiod [2]. Flower bud initiation usually starts before dormancy induction in ephemerals plants and continues afterward [4]. Several studies provided significant insights into the growth cycle, propagation, morphological attributes, and environmental impacts in E. japonicum [1,[5][6][7][8][9]. However, there is an apparent gap in research addressing the regulatory basis of underground floral differentiation in E. japonicum.
Floral organ development is critical for ornamental plants to meet commercial requirements [10,11]. During the past few decades, floriculture industry has been expanded with the inclusion of wildflowers and their domestication for commercial use [12]. Aside from commercial use, flower development is also critical for plant ecology and evolution [13,14]. The role of metabolites in pigmentation has been attributed to the variable accumulation of flavonoids, carotenoids, betalains, and chlorophylls [15]. However, metabolic changes associated with active compounds during the floral developmental stage have not been studied well. Metabolomics regulated downstream of genomics and proteomics in system biology provides tools for identifying differentially accumulated metabolites during a specific developmental stage [16]. Several reports suggested distinct metabolic profiles associated with each genotype [17,18]. Furthermore, differential accumulation of metabolites between different growth stages coupled with transcriptome can provide potential insights into the developmental process.
The shift from vegetative to reproductive stage comprised a set of fine-tuned processes regulated through developmental signals and environmental cues such as light and temperature [19]. The flowering originates from floral primordium followed by stamens and pistil differentiation which are the critical phases in flowers development [20]. Therefore, understanding the regulatory mechanisms underlying stage-specific differentiation is essential. Floral differentiation has gained much attention in plant species. For instance, Jatropha curcas [20], Brassica napus [21], Camellia sinensis [22], Populus [23], Dianthus caryophyllus [24], Litsea cubeba [25], Rosa chinensis [26], Lilium [27], and Juglans regia [28] have been well studied for molecular mechanisms underlying reproduction phase. Previous statistics suggested the regulatory role of anthocyanins, gibberellic acid, flavonoids, vernalization, and developmental pathways during floral differentiation [29][30][31]. Florigen, encoding a conserved protein FLOWERING LOCUS T, is a systematic switch for flowering control [32]. Moreover, FT regulation is mediated by GIGANTEA-CONSTANS-FT complex under strict photoperiod sensitivity [22,32]. The complex regulatory network of flowering is activated by external variables, such as day length and humidity. The environmental variables play a crucial during floral differentiation [33]. However, an overlap between flowering and dormancy regulation pathways complicates understanding specific regulatory pathways [33]. Studies have reported simultaneous reduced expression of FL and FRIGIDA during vernalization [34,35]. Understanding the overlap between flowering and dormancy can provide mechanistic insight into flowering regulation during dormancy period.
In this study, we profiled the metabolome and transcriptome of E. japonicum at four floral development stages (flower primordium differentiation, perianth differentiation, stamen differentiation, and the pistil differentiation period) to understand the molecular changes underlying the underground floral differentiation.

Plant
Material and Sampling. The study area includes the Tuodaoling region in Northeast China, with a high occurrence of Erythronium japonicum Decne (Liliaceae) populations. Plant samples were collected to the protocol described in our previous study [36]. Samples were collected in three replicates at four different stages, including flower primordium differentiation, perianth differentiation, stamen differentiation, and the pistil differentiation period. Cryopreserved samples were used for further downstream analysis following the methods of Wang et al. [36] and Gao et al. [37].
2.3. Quality Check. Quality check for multistage metabolome was performed according to Fiehn et al. [40]. Descriptive statistics for each dataset were obtained using Analyst 1.6.3 software (AB Sciex, Ontario, ON, Canada). Moreover, variability of datasets was estimated using principal component analysis (PCA) and Pearson correlation coefficient (PCC) with prcomp within R (http://www.r-project.org/).

Metabolic Profiling.
Metabolites are generally considered as the bridge between molecular mechanisms and phenotype. Therefore, to understand the molecular characteristics underlying flowering in E. japonicum, we performed systematic metabolic profiling based on widely targeted metabolomics (UPLC-MS/MS) at four floral differentiation stages, including flower primordium differentiation (Az), perianth differentiation (Bz), stamen differentiation (Cz), and the pistil differentiation period (Dz). The quality was confirmed by monitoring the instrument's accuracy following qualitycontrol measures, as formerly described by Fiehn et al. [40]. Superimposed display analysis of mass spectrometry total ion current (TIC) and extracted-ion chromatogram (XIC) of samples which were run at a different time ( Figure S1). The overlapped TIC suggested the stability of the instrument as a quality check.
The results were further verified using PCC and PCA for all the samples (Figures 1(c) and 1(d)). A strong correlation within replicates of each sample was observed, conferring the consistency of metabolome datasets used in the study. Metabolome from different tissues depicted a relatively weak correlation between different samples. Moreover, a scatter plot based on the first two PCs clustered samples into four groups consisting of replicates from each sample. PCA results, depicting 58.41 variations with PC1 (37.75%) and PC2 (20.66%), also verified the reliability of the metabolome dataset.
We compared the metabolic profiles to clarify the differentially regulated metabolite accumulation between four floral developmental stages ( Figure 2). The comparative metabolic profile suggested differential regulation of metabolites between different developmental stages. One hundred sixty-two metabolites were found with differential regulation between primordium differentiation (Az) and perianth differentiation (Bz). Among 160 DAMs, 119 were upaccumulated, and 41 depicted downaccumulation (Table S2) Table S3-S7). The results suggested a significantly higher number of DAMs when the primordium differentiation stage was compared with later stages. However, the number of DAMs reduced when later stages were compared. Furthermore, we identified 100 conserved DAMs when the flower primordium differentiation (Az) stage was compared with perianth differentiation (Bz), stamen differentiation (Cz), and the pistil differentiation period (Dz). KEGG enrichment analysis showed that these pathways are related to the cellular differentiation and transition from vegetative to reproductive stage, suggesting a significant role of DAMs in floral differentiation concerning E. japonicum.

Conjoint Analysis of Metabolome and Transcriptome.
Metabolome data sets of E. japonicum were further exploited using conjoint analysis of metabolic and transcriptomic profiles of multistage floral development, including flower primordium differentiation (AZ), perianth differentiation (BZ), stamen differentiation (CZ), and the pistil differentiation period (DZ). The transcriptome data sets used in this study resulted in the identification of 9,383, 6,979, 16,758,9,522,7,387

BioMed Research International
As presented in Figure 4, comparative transcriptome and metabolome between flower primordium differentiation (AZ) and perianth differentiation (BZ) stage suggested KEGG enrichment of DEGs and DAMs in multiple regulatory pathways, including flavonoid biosynthesis, photosynthesis, and lipid metabolism (Figure 4(a)). The PCC for DAMs and DEGs was calculated using R and PCC with values higher than 0.8 are presented as a ninequadrant diagram (Figure 4(b)). Quadrants 3 and 7 repre-sented the DEGs and DAMs with consistent regulation (positively correlated). In contrast, quadrants 1, 2, 4, 6, 8, and 9 represented DEGs and DAMs with negative correlations. DEGs and DAMs with PCC greater than 0.8 were selected, and their expression pattern was presented as a heat map (Figure 4(c)). The clustered heat map depicted DAMs highly correlated with DEGs into 12 major classes, with the most abundant classes as flavonoids, lipids, and phenolics.   BioMed Research International Correlation analysis of Cluster-35905.71088; EARLY FLOWERING 3 (ELF 3) with DAMs identified at the perianth differentiation stage suggested a significantly higher correlation between ELF 3 and DAMs ( Figure 5). Lipids were among the most abundant metabolite class significantly correlated with ELF3, suggesting a substantial role of lipids in flower morphogenesis. All lipids showed a positive correlation with ELF3 except 1-O-caffeoylglycerol and choline alfoscerate, which showed a significant negative correlation with ELF 3. Three flavonoids, including apigenin-7-O-(6"p-coumaryl)glucoside, apigenin-7-O-rutinoside (isorhoifolin), and apigenin-5-O-glucoside depicted a positive correlation with ELF 3, while dihydrocharcone-4'-O-glucoside was found with a significant negative correlation. Among 15 phenolics, 13 expressed a positive correlation with ELF 3. Flowering locus K, cullin-1, PHD finger-containing protein, and ZSWIM3 (zinc finger SWIM domain-containing protein 3) also positively correlated with lipid accumulation.

Discussion
Flower development is a systematic process under strict genetic control in higher plants. It can be divided into major phases: floral induction, meristem formation, and floral organ development [41]. The complex flowering mechanism is controlled by highly conserved genes regulating transcription factors and their protein products composing a gene regulatory network (GRN). Flowering-time genes are at the top of GRN hierarchy and play a crucial role in the developmental shift. Flowering control genes are generally triggered by external factors such as photoperiod, temperature, and humidity [42,43]. However, the initial flowering phase in E. japonicum starts underground in the absence of light and vernalization [2]. Therefore, it is valuable to explore and identify the genetic regulators of floral differentiation in E. japonicum. The present study aimed at providing a systematic metabolic insight coupled with a conjoint analysis of the metabolome and transcriptome at four floral developmental stages in E. japonicum viz., primordium differentiation, perianth differentiation, stamen differentiation, and pistil differentiation stage.
Flowering time is important in securing seed production and, therefore, ensuring species survival. In addition to the gene regulatory network controlling floral differentiation, there is a significant lack of studies concerning the role of metabolites during flower development stages. Multistage metabolome comparison yielded significant variation in metabolite accumulation. For instance, a comparison of primordium differentiation (Az) and perianth differentiation (Bz) stages identified upaccumulation of p-coumaroylputrescine, scopoletin, isorhoifolin, cosmosiin, genistin, and LysoPC BioMed Research International 15 : 0. p-Coumaroylputrescine has been previously identified with a substantial role during flower development in tobacco [52][53][54]. Scopoletin plays a significant role in physiological activities [55]. Therefore upaccumulation of scopoletin and its glycosides during the perianth differentiation suggested a considerable role of scopoletin in transition from primordium differentiation to perianth differentiation. Although cosmosiin, genistin, and LysoPC 15 : 0 have been identified at the flowering stage in multiple species, their roles during the floral differentiation stages are still unknown. Furthermore, gene ontology annotation for the DMAs during the perianth differentiation stage suggested enrichment of purine metabolism and phenylpropanoid biosynthesis. The phenylpropanoid biosynthetic pathway is crucial for flavonoid pigments synthesis during flower development [56,57]. A multifaceted gene regulatory network constituting a hierarchy of coordinated gene functions is crucial for flower development. Most of the genes involved in the floral gene regulatory network encode transcription factors, such as MADS-domain, LEAFY-(LFY-) like, and APETALA2-(AP2-) like proteins [41]. Transcriptome analysis of four floral differentiation stages of E. japonicum suggested a differential expression of ELF3 and FT, cullin 1, GLP1, and CONSTANS [36]. We performed the conjoint analysis of metabolome and transcriptome to understand the coregulation of DEGs and DAMs. ELF3 expression was significantly correlated with lipids and flavonoids. The ELF3 gene, also known as the clock gene, regulates the evening complex (EC) with ELF4 and LUX leading to flowering [58,59]. Furthermore, Deng et al. characterized CONSTANS generegulating lipid biosynthesis during flower development in Chlamydomonas reinhardtii [60]. Flowering locus K, cullin-1, PHD finger-containing protein, and ZSWIM3 (zinc finger SWIM domain-containing protein 3) also depicted a positive correlation with lipid accumulation. Further molecular characterization of identified DEGs and DAMs can yield potential insights into gene regulatory network of flowering and metabolite regulation during floral development.
In sum, this study utilized the metabolome and transcriptome of E. japonicum at four underground flower developmental stages. Comparative metabolomics suggested a differential accumulation of 586 metabolites during floral differentiation. The identified DAMs were further narrowed down based on their accumulation pattern. A conjoint analysis of metabolome and transcriptome yielded insight into coregulation of DEGs and DAMs, suggesting a close link between regulation of flowering-related genes and metabolite accumulation. Further molecular insights can potentially highlight the role of metabolites during floral differentiation in E. japonicum.

Data Availability
The raw RNA-seq data have been submitted to NCBI SRA under the project number: PRJNA730644.

Supplementary Materials
Figure S1: quality check for randomly selected sample. (a, b) Integral correlation chart of quantitative metabolite analysis. The x-axis represents retention time, the y-axis represents ion current intensity (PS), and peak represent relative content of the metabolite. (c) Total ion current (TIC) over time for sample 20. Figure Table S1: metabolic profile of E. japonicum floral developmental stages. Table S2: differentially accumulated metabolites (DAMs) between flower primordium differentiation (Az) and perianth differentiation stage (Bz). Table S3: differentially accumulated metabolites (DAMs) between flower primordium differentiation (Az) and stamen differentiation (Cz). Table S4: differentially accumulated metabolites (DAMs) between flower primordium differentiation (Az) and pistil differentiation period (Dz). Table S5: differentially accumulated metabolites (DAMs) between perianth differentiation stage (Bz) and stamen differentiation (Cz). Table S6: differentially accumulated metabolites (DAMs) between perianth differentiation stage (Bz) and pistil differentiation period (Dz). Table S7: differentially accumulated metabolites (DAMs) between stamen differentiation (Cz) and pistil differentiation period (Dz). Table S8: correlation of five selected DEGs (related to floral differentiation) with DAMs identified from comparison Az vs. Bz. (Supplementary Materials)