Comparative transcriptome analysis of flower bud transition and functional characterization of EjAGL17 involved in regulating floral initiation in loquat

Floral initiation plays a critical role for reproductive success in plants, especially fruit trees. However, little information is known on the mechanism of the initiation in loquat (Eriobotrya japonica Lindl.). Here, we used transcriptomic, expression and functional analysis to investigate the candidate genes in floral initiation in loquat. Comparative transcriptome analysis showed differentially expressed genes (DEGs) were mainly enriched in the metabolic pathways of plant hormone signal transduction. The DEGs were mainly involved in the gibberellin, auxin, cytokinin, abscisic acid, salicylic acid and ethylene signaling pathways. Meanwhile, some transcription factors, including MADS-box (MCM1, AGAMOUS, DEFICIENS and SRF), MYB (Myeloblastosis), TCP (TEOSINTE BRANCHED 1, CYCLOIDEA and PCF1), WOX (WUSCHEL-related homeobox) and WRKY (WRKY DNA-binding protein), were significantly differentially expressed. Among these key DEGs, we confirmed that an AGL17 ortholog EjAGL17 was significantly upregulated at the flower bud transition stage. Phylogenetic tree analysis revealed that EjAGL17 was grouped into an AGL17 clade of MADS-box transcription factors. Protein sequence alignment showed that EjAGL17 included a distinctive C-terminal domain. Subcellular localization of EjAGL17 was found only in the nucleus. Expression levels of EjAGL17 reached the highest at the development stage of flower bud transition. Moreover, ectopic expression of EjAGL17 in Arabidopsis significantly exhibited early flowering. Our study provides abundant resources of candidate genes for studying the mechanisms underlying the floral initiation in loquat and other Rosaceae species.


Introduction
Floral initiation is a critical phase for reproductive success in the life cycle of plants [1,2]. Molecular mechanisms of floral initiation are based primarily on studies in the model plant Arabidopsis [2][3][4]. Flower initiation is controlled by the environmental signals and molecular networks in Arabidopsis. Among the molecular networks, floral integrators including FLOW-ERING LOCUS T (FT) and SUPPRESSOR OF OVEREXPRESSION OF CONSTANS 1 (SOC1), and transcription factors such as MADS-box (MCM1, AGAMOUS, DEFICIENS and SRF) genes play important regulatory roles in floral initiation and flowering time control, providing a fundament for studying gene regulation of flower development in angiosperm [5][6][7][8][9][10][11][12][13].
Transcriptomics is an effective method for discovering the gene networks and regulatory mechanisms underlying floral initiation in non-model plants [14][15][16]. For example, in Rosa chinensis, comparative transcriptome analysis of the floral transition revealed that differentially expressed genes (DEGs) are mainly involved in the pathways of vernalization, photoperiod and gibberellin (GA) [14]. In Agropyron cristatum, transcriptomic analysis of floral transition indicated that DEGs are mainly involved in plant circadian clock pathway and flowering timeassociated genes [15]. Comparative transcriptome analysis indicated that DEGs of cytokinin biosynthesis pathway are mainly associated with floral transition in Dendrobium nobile. However, global gene expression changes of floral initiation in Eriobotrya remain unknown and need further investigated.
Loquat (Eriobotrya japonica Lindl.) is a subtropical evergreen species and belongs to the Amygdaloideae subfamily of the Rosaceae [17]. Floral initiation of loquat is a continuous process, which is not interrupted by winter dormancy [18,19]. In this study, we first investigated the transcriptomic changes in the floral transition of loquat between vegetative apex (VA) and flower bud transition (FBT). The DEGs were mainly involved in the metabolic pathways of hormone signal transduction. Furthermore, we identified that a candidate gene EjAGL17 was significantly upregulated at the flower bud transition stage and involved in floral initiation. Our results provide abundant gene resources for studying the mechanisms underlying the floral initiation in loquat.

Plant material
Flower buds of loquat cultivar 'Changbai No.1' at different development stages were provided and collected from the experimental farm of Southwest University in Xiema town (Chongqing, China). The field experiments were conducted under local legislation and permissions. According to our previous study, the development stages included vegetative apex (S1), flower bud transition and flower bud differentiation (S2), elongation of the main axis in a panicle (S3), rapid panicle elongation (S4), visible floral buds (S5), elongation of branches in a panicle (S6), white corollas (S7) and full bloom (S8) [20]. Then, these samples were frozen immediately in liquid nitrogen and stored at -80˚C until used.

Morphological analysis of the floral initiation
Morphology of developmental stages was collected and photographed using Canon EOS 6D (Canon Inc., Tokyo, Japan) according to the stages described previously [20]. The buds of loquat at the stage of vegetative apex (VA) and flower bud transition (FBT) were fixed in a FAA solution under vacuum. Samples were dehydrated by a series of ethanol solution, and then decolorized by dimethylbenzene: ethanol (50:50, by vol.) and 100% dimethylbenzene. Then, samples were treated using three changes of 100% paraffin at 60˚C and finally embedded. The paraffin-embedded samples were cut to a thickness of 10 μm. After dewaxing, the samples were stained with safranine-fast green solution and photographed using an Observer DP80 microscope (Olympus, Tokyo, Japan).

RNA extraction, library consatruction and sequencing
Total RNA was extracted from VA and FBT samples using the EASYspin Plant RNA Extraction kit (Aidlab, Beijing, China). Three independent biological replicates were performed for each sample. Residual DNA was removed from the total RNA using RNase-free DNase I (Takara, Japan) at 37˚C for 20 min. Integrity of RNA was validated by 1% agarose gel and Agilent 2100 Bioanalyzer (Agilent Technologies, Palo Alto, CA, USA). Then, oligo (dT)-attached magnetic beads enriched method was used in purifying mRNA.
The mRNAs were broken into short fragments and synthesized second-strand cDNA. Then, the synthesized cDNA fragments were subjected to end pairing, the addition of a single 'A' nucleotide and ligation with sequencing adapters. The ligation fragments were amplified by PCR, and sequenced using an Illumina HiSeq X-Ten platform (Shanghai Personal Biotechnology Co., Ltd., China). Sequencing raw data are available with an accession number CRA003214 in the Genome Sequence Archive (GSA, https://bigd.big.ac.cn/gsa/).

Assembly and functional annotation
The raw reads containing adapter sequences and low quality bases were removed. Reads containing ambiguous 'N' bases, and fragment less than 50 bp were filtered to generate high quality clean reads. Then, the clean reads were subsequently assembled using Trinity assembly program [21]. Basic Local Alignment Search Tool (BLAST) was used to annotate the function of unique sequences using the databases of the non-redundant (NR) proteins, Swiss-Prot, Gene Ontology (GO) [22], evolutionary genealogy of genes: Non-supervised Orthologous Groups (eggNOG) [23] and Kyoto Encyclopedia of Genes and Genomes (KEGG) [24].

Analysis of differentially expressed genes (DEGs)
Expression of unigenes were calculated by the method of fragments per kb per million from the mapped reads (FPKM) [25]. Differential expression of genes were analyzed using DESeq software [26], and unigenes showing at least a two-fold change and P-value � 0.05 were considered to be DEGs. Then, the DEGs were mapped to GO terms and KEGG pathways the method described previously [27,28].

Validation of RNA-Seq analysis by qRT-PCR
Total RNA was extracted from the collected from VA and FBT. Then, 2 μg RNA of each sample was used for cDNA synthesis using the PrimeScript RT reagent kit (Takara, Japan). For quantitative real-time PCR (qRT-PCR), 1 μL cDNA was used as a template in a 20 μL PCR. Then, qRT-PCR was performed using CFX96 Touch Real-time PCR Detection System (BioRad, USA). The parameters were at 95˚C for 5 min, followed by 40 cycles of 95˚C for 20 s, 56˚C for 20 s, 72˚C for 20 s; then, temperature change 1.0˚C/s to 95˚C for melt curve. The actin of loquat was used as a normalization control with the primers qEjactinF and qEjactinR [29]. These primers of subcellular localization and qRT-PCR were shown in S1 Table. Three biological replicates of each sample were performed. The expression levels were calculated using the 2 -ΔΔCT method [30].

Isolation of an AGL17 ortholog in loquat
Total RNA was extracted from flower buds of loquat as described above. The extracted RNA was treated with DNase I (TakaRa, Japan), and the concentrations were detected by UV spectrophotometry (NanoDrop 2000, Thermo Scientific, USA). Then, the DNase I-treated RNA was used for rapid amplification of cDNA ends (RACE). From these key DEGs obtained by RNA-Seq, the primers of an AGL17 ortholog (EjAGL17) were designed from the DN37075_ c0_g1 sequence. The 3 0 RACE of EjAGL17 was performed using 3 0 -full RACE Core Set Version 2.0 kit (Takara, Japan) and 2 μg of the DNase I-treated RNA. First PCR was performed using 3 0 RACE Outer Primer and the gene-specific primer 3RAGL17F1. A second PCR was performed using 3 0 RACE Inner Primer and 3RAGL17F2. PCR parameters were a 5 min at 95˚C denaturation step, followed by 30 cycles of 40 s at 95˚C, 40 s annealing at 56˚C and 40 s extension at 72˚C, with a final step of 10 min at 72˚C. The 5 0 RACE of EjAGL17 was carried out using 2 μg of the DNase I-treated RNA and SMARTer RACE 5 0 /3 0 Kit (Takara, Japan). PCR amplification was performed using Universal Primer Mix, the gene-specific primer 5RAGL17R1 and 5RAGL17R2. PCR parameters were 5 cycles at 95˚C for 30 s, 72˚C for 3 min; 5 cycles of 95˚C for 30 s, 70˚C for 30 s, 72˚C for 3 min; 25 cycles of 95˚C for 30 s, 68˚C for 30 s and 72˚C for 3 min.
2 μg of the DNase I-treated RNA was used for cDNA synthesis using M-MLV reverse transcriptase and oligo(dT)-15 adaptor primer (Takara, Japan). Then, the cDNA sequence integrity of EjAGL17 was verified using the primers FLAGL17F and FLAGL17R. PCR parameters were at 95˚C denaturation step for 5 min, followed by 30 cycles of 95˚C for 50 s, 56˚C for 50 s and 72˚C for 40 s, with a final extension step of 72˚C for 10 min. The primer sequences of EjAGL17 isolation were shown in S2 Table. Finally, PCR products were recycled and cloned into pMD18-T vector (Takara, Japan) and sequenced.

Analyses of sequence alignment and phylogenetic tree
The BLAST analysis of amino acid sequences of EjAGL17 was performed. Multiple AG, AGL17, SOC1, AP1, SEP, AP3, PI and TM6-clade proteins from various eudicots were selected for alignment from the GenBank database. Accession numbers of these clade proteins were selected for alignment from various eudicots and listed in S3 Table. A ClustalW program was used to align amino acid sequences of these proteins including the M, I, K and C domains [31]. A phylogenetic tree was constructed by MEGA 5.0 software [32,33], and the parameters were the maximum likelihood method, 1000 replicates of bootstrap and Jones-Taylor-Thornton (JTT) model [34,35].

Subcellular localization and quantitative real-time PCR
Subcellular localization of EjAGL17 was observed using the modified pCAMBIA 1300 vector and the primers SLEjAGL17F and SLEjAGL17R [36]. Then, the 35S::EjAGL17-GFP was constructed and transformed into Nicotiana benthamiana leaves using Agrobacterium-mediated transient transformation. The vector expressing GFP alone was as a negative control. Green fluorescent protein (GFP) fluorescence signals were observed using a fluorescence microscope Observer DP80 (Olympus, Japan).
Total RNA was extracted from the collected flower buds of eight developmental stages (S1-S8). Then, 2μg RNA of each sample was used for cDNA synthesis as described above. For quantitative real-time PCR (qRT-PCR), the primers of subcellular localization and qRT-PCR were shown in S4 Table.

Vectors construction and Arabidopsis transformation
The EjAGL17 coding sequences was cloned into the expression vector pBI121 (BD Biosciences Clontech, USA) using the restriction enzymes of Xba I and Sma I (Takara, Japan), and the primers TEjAGL17F and TEjAGL17R. The vector expressing EjAGL17 under the control of the CaMV 35S promoter was constructed and transformed into Agrobacterium tumefaciens strain GV3101-90. The 35S::EjAGL17 was transformed into wild-type Arabidopsis using the floral-dip method [37]. Then, transgenic seeds were germinated on a solid 0.5×MS medium including 50 μg/ml kanamycin at 4˚C for 2 days. These seeds after vernalization were transferred to the greenhouse under the conditions of 16 h light/8 h dark at 22˚C for 2 weeks. Transgenic Arabidopsis lines were selected according to the described method [35]. Subsequently, the EjAGL17 expression levels of transgenic lines were detected by qRT-PCR with the primers qEjAGL17F and qEjAGL17R, qagl17F and qagl17R [3]. As an internal control, the Actin of Arabidopsis was used with the primers qactinF and qactinR [38]. These primers of vectors construction and qRT-PCR were shown in S4 Table. 3. Results

Analysis of sequencing, assembly, annotation and DEGs
Developmental stages of vegetative apex (VA) and flower bud transition (FBT) were observed in loquat (Fig 1). The VA was embraced by rudimentary leaves. Then, FBT were formed from VA and the bud was swelled. The paraffin section observation of the apical buds showed that the transformation from vegetative to reproductive growth was from VA to FBT (Fig 1Vc and  1Fc).
To identify candidate genes for the floral initiation, RNA-Seq libraries were constructed from VA and FBT. Over 6.1 Gb data of RNA sequencing from each sample was obtained with Q20 and Q30 values that were all greater than 90% (S5 Table). A total of 109 965 unigenes were generated with an average length of 837.1 bp and N50 of 1 527 bp (Table 1). Principal component analysis (PCA) analysis of global gene expression in VA and FBT were classified into two distinct groups, reflecting significant differences in global gene expression from VA to FBT. For each sample of the two developmental stages, three biological replicates were gathered into one cluster, suggesting the correlation of the biological replicates were high (Fig 2A).
In total, 52 750, 45 770, 41 679, 18 950 and 6045 unigenes were annotated in the NR, egg-NOG, Swiss-Prot, GO and KEGG databases, respectively. In all five databases, 3379 (3.07%) unigenes were annotated (S1 Fig and S6 Table). Based on GO term analysis, the unigenes were classified into biological process, cellular component and molecular function (S2 Fig). The biological processes were mainly involved in metabolic process, cellular process and single-organism process. The cellular components were mainly cell, cell part and membrane. The molecular function was mainly involved in binding and catalytic activity. Meanwhile, the unigenes from the two stages were mapped into the pathways of metabolism, genetic information processing, environmental information processing, cellular processes and organismal systems on the basis of KEGG enrichment (S3 Fig).

Analysis of key DEGs involved in hormone signal transduction pathways
To obtain gene alterations in floral initiation, comparative transcriptome analysis of FBT vs VA were performed to detect floral initiation-related genes. In comparison of FBT vs VA, 13 087 DEGs were identified, of which 10 580 were upregulated and 2507 were downregulated ( Fig 2B). The DEGs were mainly involved in plant hormone signal transduction (23 transcripts), carbon metabolism (29 transcripts), phenylpropanoid biosynthesis (17 transcripts),  To further investigate plant hormone signal transduction of the floral initiation, we analyzed gene expression of key DEGs in the signaling pathways of gibberellin (GA), auxin, cytokinine, abscisic acid (ABA), salicylic acid (SA) and ethylene (Fig 3, Tables 2 and 3). In the GA signaling pathway, a total of 4 key genes, including GA20OX2, GA3OX1, GA2OX and KAO1, were identified and all were significantly upregulated at FBT stage ( Fig 3A and Table 2). In the auxin signaling pathway, a total of 7 key genes were identified, of which IAA3, IAA29, SAUR32, SAUR50, GH317 and GH36 were significantly upregulated at FBT stage (Fig 3A and  Table 2). A total of 2 key genes, including AHK4 and ARR3, were identified and significantly upregulated in cytokinin signaling pathway at FBT stage ( Fig 3A and Table 2). In ABA signaling pathway, six key genes were identified, including ABI5, AI5L5, PP2C8, PP2C16, PP2C24 and PYL4, of which PYL4 was significantly downregulated (Fig 3B and Table 3). In SA signaling pathway, two genes, including PRB1 and TGA4, were identified and both were significantly upregulated at FBT stage ( Fig 3B and Table 3). A total of 2 key genes, including ERF92 and

PLOS ONE
XTH23, were identified in ethylene signaling pathways, but ERF92 was significantly downregulated at FBT stage ( Fig 3B and Table 3).  Table 4). In MYB family genes, three key genes, including MYB3, MYB6 and MYB62 were identified, and expression levels of MYB3 and MYB6 were significantly downregulated ( Fig 4A and Table 4). In TCP family genes, a total of 5 key genes were identified, of which TCP2, TCP4, TCP5 and TCP13 were significantly upregulated at flower bud transition stage ( Fig 4B and Table 4). In WOX family genes, two genes, including WOX4 and WOX9, were identified and both were significantly upregulated at FBT stage ( Fig 4B and Table 4). In WRKY family genes, four key genes were identified, including WRKY7, WRKY19, WRKY70 and WRKY76, of which WRKY7 and WRKY70 were significantly downregulated (Fig 4B and Table 4).

Validation of the expression analysis of several key flower initiationrelated genes
Analysis of qRT-PCR was used to validate the expression levels of several key DEGs that were identified using RNA-Seq. A total of nine flower bud transition-related unigenes were randomly detected via qRT-PCR (Fig 5). The expression levels of these nine genes corresponded well with the FPKM values obtained by RNAseq, suggesting that the correlation coefficients between the FPKM and qRT-PCR values were high.

Sequence and phylogenetic analysis of EjAGL17
Based on the analysis of key DEGs, one candidate gene EjAGL17 (DN37075_c0_g1) was isolated from the MADS-box family. Comparative protein sequence analysis between EjAGL17 and other four AGL17 orthologous proteins showed that EjAGL17 has a 59-aa highly conserved MADS domain (1-59 aa), a 19-aa I domain (60-78 aa), an 80-aa K domain (79-158 aa) and a 65-aa completely divergent C domain (159-223 aa) (Fig 6). Meanwhile, these AGL17 orthologous proteins including EjAGL17 contained a distinctive C terminal domain (Fig 6). Molecular weight and theoretical isoelectric points of EjAGL17 protein are 25.59 kD and 9.48, respectively (S7 Table). Accession number of EjAGL17 sequence was MT344109 in the GenBank.  Protein sequence of EjAGL17 was blasted to reveal the molecular evolution with AG, SOC1, AP1, SEP, AP3, TM6 and PI clade proteins from other angiosperms. Phylogenetic tree analysis revealed that the EjAGL17 is grouped into the AGL17-clade of MADSbox transcription factors in the eudicots (Fig 7).

Subcellular localization and expression levels of EjAGL17
Subcellular localization of 35S::EjAGL17-GFP were detected and transiently expressed in epidermal cells of N. benthamiana leaf. As a negative control, fluorescence of the 35S::GFP was

PLOS ONE
Transcriptome analysis and function of EjAGL17 in floral initiation in loquat observed in both the nucleus and cytoplasm. However, fluorescence of 35S::EjAGL17-GFP was found only in the nucleus (Fig 8A). These results showed that EjAGL17 is a nuclear-localized protein and its subcellular localization pattern is consistent with that of AGL17 orthologs in soybean and maize [39,40], having subcellular localization characteristics of typical transcription factors.
During the process of flower development, expression of EjAGL17 was persisted from vegetative apex to full bloom in loquat. Meanwhile, the expression levels of EjAGL17 reached the highest at the development stage of FBT (Fig 8B).

Ectopic expression of EjAGL17 in Arabidopsis
To further investigate the function of EjAGL17, overexpression assays were conducted by ectopic expression of EjAGL17 in Arabidopsis. A total of thirty-five 35S::EjAGL17 transgenic Arabidopsis lines were obtained by screening and detecting with MS medium containing kanamycin. Flowering phenotypes of three T3 generation of 35S::EjAGL17 transgenic lines were observed. Comparison with untransformed lines, 35S::EjAGL17 transgenic plants caused early flowering under the same long-day conditions (Fig 9A).
Expression levels of EjAGL17 in transgenic lines were significantly higher than that of untransgenic Arabidopsis (Fig 9B). However, expression levels of Arabidopsis AtAGL17 gene were no significant difference between transgenic and untransgenic Arabidopsis (Fig 9C).

Discussion
Floral initiation of woody angiosperm undergoes morphological changes, and is generally controlled by the regulatory network and environmental signals [41,42]. However, reproductive development of loquat is a continuous process, and do not exhibit dormancy [18,19], indicating that expression pattern of genes controlling floral initiation might be distinctive. In this study, we used comparative transcriptome analysis to investigate gene expression changes at the stages of VA and FBT. The DEGs were mainly involved in the hormone signal transduction pathways and transcription factor genes. Among these DEGs, we identified that one candidate gene EjAGL17 was involved in floral initiation. Our study provides abundant candidate genes for studying the mechanisms underlying the floral initiation in loquat.

Expression pattern of key DEGs associated with floral initiation
In our study, the DEGs were mainly involved in the signaling pathways of GA, auxin, cytokinin, ABA, SA and ethylene. Previously, in Litchi chinensis, comparative transcriptomic analysis of floral initiation was mainly enriched in the metabolic pathways of plant hormone signal transduction, and the DEGs were mainly involved in the biosynthesis of auxin, cytokinin, jasmonic acid (JA), SA and ABA [43]. In Camellia sinensis, transcriptomic analysis of flower development indicated that DEGs were mainly associated with metabolic pathways of auxin, cytokinin, GA, ABA, ethylene, brassinosteroid, JA and SA [44]. In Malus domestica,

PLOS ONE
transcriptomic analysis of flower induction showed that the DEGs were mainly enriched in cytokinin, ABA and GA pathways [45]. However, transcriptomic analysis of floral bud initiation, floral organ differentiation and bud outgrowth revealed that DEGs were mainly involved in the signaling pathways of GA, auxin, JA and brassinosteroid in Camellia azalea [42]. In Annona squamosa, comparative transcriptomic analysis of floral transition and flower development reported that DEGs were mainly enriched in the signaling pathways of GA, auxin, cytokinin and ABA [46]. These previous studies and our data suggest that key DEGs are mainly enriched in hormone signal transduction pathways in floral initiation.

DEGs of key transcription factors that affect floral initiation
In our work, transcription factors, including MADS-box, MYB, TCP, WOX and WRKY, were significantly differentially expressed at FBT stage. Similarly, some transcription factors, such as MADS-box, TCP and MYB, were also significantly differentially expressed at flower development in double-flower loquat [28]. Especially, MADS-box transcription factors play important regulatory roles in floral transition, flower organ identity determination and flowering time control in the molecular networks in angiosperm [5][6][7]. Importantly, MIKC-type MADSdomain transcription factors exhibit a characteristic domain organization and contain a MADS (M), an intervening (I), a keratin-like (K) and a C-terminal (C) domain [47][48][49]. The designated domains of MADS-domain transcription factors provide a fundament for studying gene regulation of flower development in angiosperms [50].

Expression pattern and functional analysis of EjAGL17
In our study, EjAGL17 expression level were upregulated at the stage of floral initiation in loquat. Meanwhile, expression pattern of EjAGL17 is responsible largely for its functional specificity, and is a good predictor of function. This indicated that EjAGL17 might be associated with floral initiation in loquat. Previously, PpAGL17 were mainly detected in the flower in P. pyrifolia [51]. However, expression level of AGL17 ortholog in soybean was much higher in roots and nodules than that in other organs [39]. In rice, AGL17 ortholog was mainly expressed in root development [52,53]. In maize, AGL17 ortholog was mainly expressed in root xylem parenchyma cells [40]. These data showed that expression patterns of AGL17 orthologs exhibit diverse in development process in angiosperm.
Ectopic expression of EjAGL17 in Arabidopsis promoted early flowering, indicating function of EjAGL17 is involved in flowering time in loquat. Meanwhile, expression levels of EjAGL17 and AtAGL17 were analyzed in transgenic Arabidopsis to verify that the early-flowering phenotype was caused by overexpression of EjAGL17. Similarly, overexpression of AGL17 promoted early flowering, but mutant of AGL17 caused late flowering in Arabidopsis [3,4]. Therefore, we conclude that the functions of AGL17 orthologs in regulating flowering time between loquat and Arabidopsis were conserved. However, in soybean, overexpression of AGL17 orthologs in transgenic soybean significantly promotes the formation and development of lateral root [39]. Function of AGL17 ortholog is involved in the regulation of root and seed development in rice, and participated in responses to osmotic stress and nutrient supply [52][53][54]. In maize, functional analysis of AGL17 ortholog suggests that its involvement in root nitrate-foraging [40]. These data revealed that characterization of AGL17 orthologs is involved in regulating different development aspects and functional diversification. Furthermore, function of EjAGL17 should be further explored to reveal the possible roles in the control of other developmental programs. Our results provide a better understanding for the functional roles of AGL17 orthologs involving in floral induction in Eriobotrya.

Conclusions
In this study, transcriptome expression and functional analyses were used to analyze gene expression changes at the stages of VA and FBT. The DEGs were mainly involved in the hormone signal transduction pathways. Meanwhile, some transcription factors were significantly differentially expressed. Among these DEGs, one candidate gene EjAGL17, which was significantly up-regulated at the development stage of floral initiation, was isolated and identified from loquat. Analyses of protein sequence and phylogenetic tree revealed that EjAGL17 was grouped into the AGL17 clade MADS-box transcription factors, and included a distinctive Cterminal domain. Ectopic expression of EjAGL17 transgenic Arabidopsis significantly promoted early flowering. Furthermore, other key DEGs should be further explored to reveal their possible roles in floral initiation. Our results provide a better understanding for metabolic pathways and candidate genes involving in floral induction in Eriobotrya.