The wild type chrysanthemum cultivar ‘Jinba’ was obtained from the Chrysanthemum Germplasm Resource Preserving Centre, Nanjing Agricultural University, China. The CmPht1;2 overexpression line-Oe1 was generated by our group [12]. The cuttings of chrysanthemum cultivated in a greenhouse at 25°C, a 16-h photoperiod with an irradiance of 2000 Lux, and a relative humidity of 70%. The rooting process was cultured in deionized water for 14 days. They were then transferred to hydroponic solutions consisting of diluted Hoagland solutions (1:4) and (1:2) for 3 d [17]. The nutrient solution renewal was carried out every 3 d. After a week of curing in total nutrient solution, the plants were respectively treated in normal phosphate concentration (HP, 300 μM Pi) or low phosphate concentration (LP, 15 μM Pi) hydroponic solution for phosphorus sufficient and deficiency treatment, respectively. The roots and shoots were harvested at 0, 2, and 7 d after the phosphate treatment, snap-frozen in liquid nitrogen, and kept at −80°C. The shoots selected here were the aboveground part including shoots and leaves. Experiments included three replicates. Each replicate contained nine seedlings.

The total RNA from the above samples was extracted using Quick RNA isolation Kit (Huayueyang Biotech Co., Ltd., Beijing, China) according to the manufacturer’s instructions. First strain cDNA was carried out by SPARKscript II RT Kit (With gDNA Eraser) (Shandong Sparkjade Biotechnology Co., Ltd., China). 500 ng·μL⁻¹ RNA was used for cDNA synthesis.

After the RNA samples were qualified, eukaryotic mRNAs were enriched with Oligo(dT) magnetic beads. Then, mRNA was fragmented into short fragments by adding fragmentation buffer. The mRNA was used as template and 6 base random primer was used as random hexamers to synthesize first strand cDNA. Then buffer, dNTPs, DNA polymerase I and RNase H were added to synthesize two strands cDNA. Then last PCR amplification was performed after cDNA modification. AMPure XP beads were used to purify double strand cDNA to obtain the final product Library. After library construction, Qubit2.0 was first used for preliminary quantification, and the library was diluted to 1.5 ng·uL⁻¹. Agilent 2100 was then used to detect the insert size of the library. The qPCR method was used after the insert size met the expectation. The effective concentration of the library was accurately quantified (Library effective concentration >2 nM) to ensure the quality of the library. The standard Illumina protocol was followed thereafter to develop mRNA-seq libraries. Subsequently, the library preparations were sequenced on an Illumina HiSeq2000 platform at Beijing Novogene Bioinformatics Technology Corporation and paired-end reads were generated. The screening criteria for the differential genes were: padj <0.05.

The verification of differential expression genes was performed using SYBR Green I Real Time PCR. 2 × SYBR qPCR Mix (5.0 μL, Shandong Sparkjade Biotechnology Co., Ltd., China), forward/reverse primer (10 μM, 0.5 μL each), cDNA template (1.0 μL) and ddH₂O (3.0 μL) were added into a 384-well plate. The reference sequence, a 151 bp fragment of EF-1α, which amplified with the primers EF1α-F/-R and the specific primers of c137909_g1 and c183919_g1 were shown in Supplementary Table S1. The experiment was performed on ROCHE (LightCycler® 480 II, Switzerland) real time fluorescence quantitative PCR instrument with 94°C 2 min; 94°C 20 s, 55°C 20 s, 72°C 30 s, 40 cycles. Finally, add the dissociation curve stage. Relative transcription levels were estimated using the 2^−ΔΔCt method [18].

The genome-wide research was investigated in the roots and shoots of Oe1 and WT under LP treatment for 0, 2, and 7 d. The reads from roots or shoots were pooled together for more comprehensive reconstruction of transcripts. In roots, 617,681 unigenes was compared and annotated in seven functional databases: 230,009 (NCBI non-redundant protein sequences (NR): 37.23%), 72,935 (NCBI nucleotide sequences (NT): 11.80%), 110,894 (euKaryotic Ortholog (KO): 17.95%), 234,503 (A manually annotated and reviewed protein sequence database (SwissProt): 37.96%), 271,994 (Protein family (PFAM): 44.03%), 276,138 (GO: 44.70%) and 144,621 (euKaryotic Ortholog Groups (KOG): 23.41%). In shoots, 207,271 unigenes were compared and annotated: 81,214 (NR: 39.18%), 40,853 (NT: 19.70%), 37,988 (KO: 18.32%), 72,736 (SwissProt: 35.09%), 74,950 (PFAM: 36.16%), 76,557 (GO: 36.93%) and 37,906 (KOG: 18.28%) (Table 1).

GO assignments were used to predict the functions of unigenes obtained above by classifying them into various biological processes. The unigenes annotated in the GO database were categorized into 56 functional groups which belong to Biological processes, Cellular component, and Molecular function. Among them, there were 25 subclasses of biological processes, 21 cellular components and 10 molecular functions. Most of these unigenes were classified into the cellular process, metabolic process and binding both in roots and shoots (Fig. 1). The unigenes were aligned to the 26 KOG classifications to classify potential functions in roots, while 25 KOG classifications in shoots. Both in roots and shoots, assignments to (J) translation, ribosomal structure and biogenesis, (O) posttranslational modification, protein turnover, and chaperones, and (R) general function prediction were all over 10.0% (Fig. 2). (T) Signal transduction mechanism was also accounted for a large proportion in roots rather than shoots (Fig. 2A). To further identify the biological pathways of Oe1 and WT, the unigenes were annotated by KEGG database. The unigenes were mainly focused on translation, folding sorting and degradation, carbohydrate metabolism, transport and catabolism, and amino acid metabolism both in roots and shoots (Fig. 3).

The significant DEGs were then identified with the applied criteria (p < 0.05) (Fig. 4). In roots, 94 DEGs (88 up-regulated genes and 6 down regulated genes) were identified between Oe1 and WT at 0 d (Fig. 4A). After 2 d LP-treatment, the number of DEGs (450 DEGs, 359 up-regulated genes and 91 down regulated genes) was increased immediately (Fig. 4B). However, only 27 DEGs (20 up-regulated genes and 7 down regulated genes) were identified between Oe1 and WT at 7 d (Fig. 4C). Meanwhile, in shoots, only 19 genes (11 up-regulated genes and 8 down-regulated gene) and 9 genes (2 up-regulated genes and 7 down-regulated gene) showed significant difference in expression levels in 0 and 2 d, respectively (Figs. 4D, 4E). In addition, reach up to 1,787 DEGs (716 up-regulated genes and 1,071 down regulated genes) were identified between Oe1 and WT at 7 d (Fig. 4F).

The influence of LP stress on biological pathways was evaluated via enrichment analysis of DEGs in KEGG between Oe1 and WT. The DEGs were assigned into 14 pathways in 0 and 7 d in roots, respectively. However, only 1 or 2 DEGs were found in every pathway (Figs. 5A, 5C). After 2 d LP-treatment, the DEGs were assigned into 63 pathways in roots. We selected the top 20 pathways that were most significant in the enrichment process to be shown in this diagram (Fig. 5B). Among these pathways, the ribosome (37 DEGs), phenylpropanoid biosynthesis (6 DEGs), Glycolysis/Gluconeogenesis (6 DEGs), oxidative phosphorylation (6 DEGs) and plant-pathogen interaction (5 DEGs) pathway showed a significant degree of enrichment by DEGs. In addition, it was important to note that phenylpropanoid biosynthesis (6 DEGs) and flavonoid biosynthesis (2 DEGs) showed a significant degree of enrichment by DEGs in roots (Supplementary Table S2).

In shoots, only 5 pathways were enriched in 0 d (Fig. 5D). While in 2 d, there was not any pathway enriched by KEGG. Suffering from phosphorus deficiency for 7 d, the DEGs were assigned into 74 pathways in shoots (Fig. 5E). Among the top 20 pathways, the starch and sucrose metabolism (36 DEGs), photosynthesis (31 DEGs) and Photosynthesis-antenna proteins (21 DEGs) showed a significant degree of enrichment by DEGs. Phenylpropanoid biosynthesis (17 DEGs) and flavonoid biosynthesis (5 DEGs) also showed a significant degree of enrichment by DEGs in shoots (Supplementary Table S3).

As the remarkable differences at the metabolomics level, pyruvate metabolism and TCA cycle pathways had mainly discussed in our previous research. Here, these two pathways were also showed difference after 2 d LP-treatment in roots between Oe1 and WT. Part of the genes were identified in these key metabolic pathways. c137909_g1 in pyruvate metabolism and c183919_g1 in TCA cycle were up-regulated 18.25- and 3.62-folds more than those of WT under LP stress, respectively (Fig. 6).

Multi-transcriptome database and candidate genes were provided for further study of genotypes with and without low-phosphorus tolerance in maize. The P deficiency-responsive genes common to low P-tolerant and the low P-sensitive line were associated with acid phosphatase (APase) activity and a series of metabolic pathways [15]. Transcriptome analysis reveals candidate genes related to low-P stress in sorghum. GO enrichment analyses showed that the candidate genes were involved in oxidoreductase activity. These will form the molecular network of sorghum responding to low P stress, and then contribute to the high-quality breeding with low P tolerance [19]. GO analysis showed that apple trees responded to metabolic processes, cell proliferation, biological process regulation, reactive oxygen metabolism and flavonoid metabolism under phosphorus stress. Pathway analysis in KEGG further showed that DEGs acted on mitogen-activated protein kinase (MAPK) signaling pathway, flavonoid biosynthesis, phenylpropanoid biosynthesis, and ATP binding box (ABC) transporters [20].

The majority of unigenes excavated in the taproot of Radish were predominately involved in basic physiological and metabolic processes, catalytic, binding, and cellular processes [21]. Gene ontology analysis revealed that salt stress-related categories involved in catalytic activity, binding, metabolic processes and cellular processes of the prunellae spica transcriptome under salt stress [22]. Similarly, here in our work, by the GO database, cellular process, metabolic process and binding were in the majority. These processes may indicate the endurance to abiotic stress of chrysanthemum. Meanwhile these unigenes were mainly annotated as translation and posttranslational modification both in roots and shoots by KOG, which contribute on the production of proteins. Nearly 200 different types of Post-translational modifications (PTMs) have been identified in previous studies, and most of them are critical regulators of protein function [23]. Under low phosphorus conditions, plant roots receive stress signals and transport them to above-ground tissues [24]. Similar to the results of the difference of MAPK signaling pathway genes in apple under LP stress, signal transduction mechanism performed noticeably in roots than shoots here [20]. The numbers of DEGs and difference pathways in roots were significantly higher than shoots under 2 d LP-treatment. While these numbers in shoots were significantly higher than those in roots under 7 d LP-treatment. This change might owe to the difference of signal transduction mechanism in roots which transmitted low phosphorus signal from roots to shoots. Carbohydrate metabolism, transport and catabolism were specially separated by KEGG in our data. These results checked with the data on many plant stress in several species. Carbohydrate metabolism focuses on biosynthesis, catabolism and biological function of the major storage carbohydrates during growth and development of the plant [25]. Soluble sugars, an important part of carbohydrate metabolism, provide important protection to plant stress. The accumulation of sorbitol and sucrose in the apple roots was higher, and the activities of sucrose synthase, invertase and sorbitol dehydrogenase, which are involved in the degradation of sucrose and sorbitol, were significantly increased under a low nitrogen supply [26]. However, the influence by the overexpression of phosphate transporter genes had rarely reported before. Therefore CmPht1;2-overexpression or high phosphate concentration of Oe1 might make an effect on the biological processes after LP treatment both in roots and shoots [12,27].

TCA cycle is one of the iconic pathways in terms of energy metabolism, being responsible for the oxidation of respiratory substrates to drive ATP synthesis [28]. Mitochondrial TCA cycle signalling involved in the regulation of DNA and histone modifications to enable dynamic control of expanded genome during eukaryogenesis. And TCA cycle remodeling is also used to coordinate stress responses and gene expression programmes in eukaryotic organisms [29]. Intermediates of glycolytic and TCA cycle were crucial players response to oxidative stress in Arabidopsis [30]. Pyruvate is the key node connecting glycolysis and TCA cycle. Genome-wide association study in wheat indicated that 6 genes involved in pyruvate metabolism and TCA cycle were upregulated significantly in the P efficient variety under LP stress [31]. D-lactate is the end product of glyoxalase system. D-Lactate dehydrogenase (D-LDH) catalyzes the oxidation of D-lactate to pyruvate using flavin adenine dinucleotide as a cofactor in Arabidopsis [32]. D-LDH completes the detoxification process of methylglyoxal and diverts the stress induced toxic metabolites D-lactate towards energy production in rice [33]. Thereby D-LDH could protect the cells from their deteriorating effects and enhance resistance to multiple abiotic stresses of plants. Here. in our results, c137909_g1 in pyruvate metabolism which was commented to D-LDH had a significantly higher expression in the Oe1 roots compared with WT under LP condition. These were able to provide a detoxification process of methylglyoxal and a further transformation from D-lactate to energy. While instead of increase, pyruvate went down as the increase of D-LDH in the metabolism data of our previous research [12]. We hypothesized that this is due to the rapid transformation of pyruvate into the TCA cycle by way of acetyl-CoA. ATP-citrate lyase (ACLY) could catalyze the transformation among acetyl-CoA, oxaloacetate and citrate based on the KEGG database. So ACLY is the key node of pyruvate metabolism going into the TCA cycle. Here c183919_g1 (ACLY) also showed a higher expression in the Oe1 roots than WT and perhaps play a key role in promoting the TCA cycle. Malate affected the length of the primary root and the number of tips in sorghum suffering from low-P stress [19]. Root-exuded organic acids will simultaneously and disproportionately increase to facilitate the Pi mobility under P deficiency [34]. In the early work, we found a prominent decrease of malate, fumarate and succinate in TCA cycle, and speculated an enhanced secretion of organic acids into the rhizosphere [12]. Therefore, the up-regulation of D-LDH and ACLY in the Oe1 plants compared with WT might enhance the growth potential and root development through promoting the pyruvate metabolism and TCA cycle in chrysanthemum. The general phenylpropanoid metabolism can generate an enormous array of secondary metabolites to result in an organ and developmentally specific pattern of metabolites in plants [35]. Salt interferences to metabolite accumulation and flavonoid biosynthesis in Tetrastigma hemsleyanum might reveal the response to salt stress by flavonoid biosynthesis in plants [36]. Here, phenylpropanoid biosynthesis and flavonoid biosynthesis were shared by both roots and shoots, but with different number of DEGs. These results might indicate the different organ and development specific pattern between roots and shoots, and the resistance response to LP stress in chrysanthemum.