Integrity transcriptome and proteome analyses provide new insights into the mechanisms regulating peel cracking in Akebia trifoliata fruit

used for of each gene control gene novo qPCR those gene sequences against NCBI The contained 12.5 µL SYBR Green PCR master mix, 1 µL cDNA, and 0.5 µL of each primer in a nal reaction volume of 25 µL. The thermal cycling program began with 3 min at 95 °C, followed by 40 cycles of 95 °C for 10 s, 55 °C for 30 s, Melt Curve 65 to 95 °C, increment 0.5 °C for 5 s. After PCR amplication, the quantitative variation was analyzed using Delta Ct method, and the analysis of statistically signicant differences from gene expression was performed by the independent samples t-test analysis at P < 0.05 using GraphPad Prism 8 software. Correlation analysis between cell wall-related gene and protein expression performed Pearson’s correlation coecient analysis galactose metabolism-associated gene and protein expression. c and i Heat map of amino sugar and nucleotide sugar metabolism-associated gene and protein expression. d and j Heat map of starch and sucrose metabolism-associated protein expression. e and k Pentose and glucuronate interconversions metabolism-associated gene and protein expression. f Cell wall related transcription factor. l cell wall metabolism-associated protein expression.

discovery of novel genes and their functions, molecular markers, and physiological stress responses in plants [23,24]. Therefore, in this study, integrative analysis of the transcriptomes and proteomes were performed to illuminate the mechanism of A. trifoliata fruit cracking at the molecular level by RNA-seq and tandem mass tag (TMT) technologies. Our comprehensive parallel analyses will provide several new and interesting insights for understanding the molecular mechanisms in fruit cracking, enhancing fruit utilization, and biomass production.

Changes in pericarp structure
The pericarp of A. trifoliata is known to crack longitudinally as it matures. The crack in the fruit and subsequent dispersal of the seeds from along the ventral suture, which is similar to the cracking seen in other legume species, so the dynamic structures of the fruit peels in different development stages were observed (Fig. 1). In the non-cracking stage (PS), the arrangement of pericarp cell and cuticle were dense, small intercellular space and distributed continuously. The fruit cracking however, was accompanied by the cells becoming thinner and bigger, a reduction in the number of cell layers, and furthermore, the arrangement of the cells was loose and poor integrity, and they began to degrade in the initial cracking stage (PM). The cells were arranged irregularly and continue to reduce cell layers, and there was atrophy degradation in the total cracking stage (PL).

Transcriptomic analysis overview
To obtain an overview of the A. trifoliata transcriptome during fruit development and ripening, three cDNA libraries (ie., PS, PM and PL) were constructed. A total of 47.05, 46.92, and 54.00 million raw sequence reads were produced from the PS, PM and PL libraries, respectively. After removing reads with indeterminate base ratios of > 10%, low-quality reads and adaptor sequences, 46. 45, 46.35, 53.46 million clean reads with the percentage of Q30 bases and GC contents of 91.58 − 93.75% and 45.94%−48.48%, were obtained respectively (Table S1). The resultant A. trifoliata transcriptome contains 241 376 transcripts, ranging from 201 to 2000 bp, and a total of 186 054 unigenes were identi ed (Table 1), and the details of size distribution of the transcripts and unigenes are shown in Fig. S1. unigenes were shown to correspond with sequences from at least one of the public databases, and 7283 unigenes annotated in all databases, resulting in the annotation of 100,329 unigenes (41.57% of all the assembled transcripts) in A. trifoliata pericarp (Table S2).
Among of these unigenes, 11 205 were identi ed as differently expressed genes (DEGs) using absolute log2 fold change > 1 with p < 0.05 during the fruit ripening. There were 779 up-regulated and 1924 downregulated unigenes in the PM compared with the PS group, which are presented in a volcano plot in Fig.   S2a; 4623 up-regulated and 1975 down-regulated in PL, compared with the PM group, and are presented in a volcano plot in Fig. S2b. There were 1904 DEGs that were co-expressed at PM_PS and PL_PM (Table 2). Functional classi cation of the identi ed DEGs To further understand the function of the identi ed DEGs, bioinformatics analysis was performed on the basis of gene functional classi cation and hierarchical cluster analysis. GO analysis indicated that most of the DEGs in the biological processes were involved in the cellular amide metabolic processes and amide metabolic processes; structural molecular activity and oxidoreductase activity were the highest portion of the DEGs in the category of molecular functions, both in PM_PS and PL_PM; cytoplasmic parts and intracellular ribonucleoprotein complexes in PM_PS cells and PL_PM cell parts, were the highest portion of the DEGs in category of cell components respectively ( Fig. 2a-2b).
Moreover, a KEGG pathway analysis was carried out to further evaluate the DEGs. In the comparison between the PM_PS and PL_PM, many DEGs were enriched in metabolic pathways, ribosomes, and biosynthesis of the secondary metabolites. Notably, most of the DEGs involved in the cell wall-related DEGs, were downregulated in the PM, compared with the PS group, but however they were upregulated in the PL compared with the PM group ( Fig. 2e-2f).
A hierarchical cluster analysis was performed to further understand the expression changes in the cell wall-related DEGs, (Fig. 3). In all, 285 cell wall related DEGs were clustered closely both in PM_PS and PL_PM group. Most of these were involved in pentose and glucuronate interconversions, phenylpropanoid pathway, galactose metabolism, starch and sucrose metabolism, amino sugar and nucleotide sugar metabolism and transcription factors ( Fig. 3a-3f).
Con rmation of fruit cell-wall related genes in A. trifoliata by Reverse transcription real-time quantitative PCR (qPCR) To validate the results of the RNA-Seq data, an expression and correlation analysis between the qPCR and the fragments per kilobase per million reads mapped (FPKM) values obtained from the RNA-seq were performed. The 20 selected genes had shown differential expression patterns in the PM_PS and PL_PM groups, and the results of the qPCR are shown in Fig. 4. Speci cally, in the PM_PS group, most of these DEGs were down-regulated, including the phenylpropanoid pathway related genes-4-coumarate-COAligase (4CL), peroxidase (PRX), and PRX2; galactose metabolism related genes-β-galactosidases (β-GAL1 and β-GAL2); amino sugar and nucleotide sugar metabolism related gene-beta-D-xylosidase (BXL), starch and sucrose metabolism related genes-cellulase (CEL), cellulose synthase-like protein (CSLG) and Glucan endo-1,3-beta-D-glucosidase (ENDOB), and the transcription factor and cell wall metabolism genes-NAC, NAC like and EXP1. While the phenylpropanoid pathway related genes-cinnamyl-alcohol dehydrogenase (CAD) and shikimate O-hydroxycinnamoyltransferase (HCT), transcription factor and cell wall metabolism genes-BHLH and dirigent protein (DIR2), and pentose and glucuronate interconversion related genes (PL, PG and PE) were signi cantly up-regulated. In the PL_PM group, most of these genes were signi cantly up-regulated in the PL_PM group, except for 4CL, CAD, β-GAL, and EXP1. Moreover, the  Table 3). Quantitative proteome analysis To understand the molecular mechanisms of peel cracking in A. trifoliata fruits, a quantitative proteomics analysis was also performed using the TMT platform and LC-MS/MS analysis during fruit development, to complement the transcriptome analysis. Accordingly, a total of 812 625 spectra, 68 151 identi ed spectra, 12 456 peptides, and 10 572 unique peptides, were found by proteomic analysis and 2839 proteins were identi ed (Table 1). In terms of protein mass distribution, proteins with molecular weight greater than 9 kDa have a wide range and good coverage, with the maximum distribution area of 10-40 kDa (Fig. S3a). The peptide quantitative analysis of the proteins showed that the protein quantity decreased with the increase of the matching peptide (Fig. S3b).
Among of these proteins, 240 were identi ed as differentially abundant proteins (DAPs) using a foldchange > 1. Moreover, a KEGG pathway analysis was carried out to further evaluate the DAPs. In the comparison between the PM_PS and PL_PM, many DAPs were enriched in two-component systems and ribosome pathways. Notably, most of the DAPs involved in cell wall-related DAPs were upregulated in both the PM_PS and PL_PM groups. While those DAPs involved in phenylpropanoid pathways and peroxisomes were downregulated in the PM_PS and PL_PM groups, respectively ( Fig. 2g-2h).
Hierarchical cluster analysis was performed to further explore the expression changes in the cell wallrelated DAPs. A total of 40 cell wall-related DAPs were clustered closely, in both the PM_PS and PL_PM groups. Most of these were involved in pentose and glucuronate interconversions, the phenylpropanoid pathway, galactose metabolism, starch and sucrose metabolism, amino sugar and nucleotide sugar metabolism, and cell wall metabolism related proteins ( Fig. 3g-l).
Con rmation of fruit cell-wall related proteins in A. trifoliata by qPCR To validate the results of the TMT data, an expression and correlation analysis between the qPCR and the FPKM values was obtained when the TMT was performed. There were 17 genes selected that had shown differential expression patterns in the PM_PS and PL_PM groups, and the results of the qPCR of these genes are shown in Fig. 6. Speci cally, in the PM_PS group, the phenylpropanoid pathway related genes (PRX, PRX3, PRX4 and PRX5), galactose metabolism related genes (β-GAL1 and β-GAL2), amino sugar and nucleotide sugar metabolism related gene (BXL), and pentose and glucuronate interconversion related genes (PG and PG2), were signi cantly down-regulated. While starch and sucrose metabolism related genes-glucan endo-1,3-beta-glucosidase (ENBG), Furostanol glycoside 26-O-beta-glucosidase (F26G), pentose and glucuronate interconversion related genes (PL and PE), and cell wall metabolism related genes (EXP1, DIR1) were signi cantly up-regulated. In the PL_PM group, most of the genes were up-regulated, except for DIR1, EXP1, BGLU33, PG, PG2, β-GAL1, and alpha/beta hydrolase (α-HY).

Comparative analysis between protein abundance and gene expression levels
To evaluate the relationships between the transcriptomic and proteomic changes during fruit pericarp cracking, the quantitative data for DEGs and DAPs was used for correlation analysis. According to the association analysis, 1904 shared DEGs and 17 shared DAPs were identi ed in the comparison between the PM_PS and PL_PM. Among the shared DEGs, 1123 were up-regulated and 781 were down-regulated in the PM_PS, whereas 808 were up-regulated and 1096 were down-regulated in PL_PM. Of the common DAPs, 9 were more abundant and 8 were less abundant in PM_PS than in PL_PM, while 8 were more abundant and 9 were less abundant in PL_PM than in the PM_PS. (Table 3). Furthermore, 14 and 4 DAPs and their corresponding DEGs were identi ed in the PM_PS and PL_PM groups, respectively. Of these, 12 DAPs (4 with increased abundance and 8 with decreased abundance) and 4 DAPs (2 with increased abundance and 2 with decreased abundance) were regulated in the same direction as their corresponding DEGs in the PM_PS and PL_PM groups, respectively ( Fig. 6a-6b). There were more DEGs than DAPs in both the PM_PS and PL_PM groups, with signi cant differences of the trends in transcript levels and protein abundance.
Furthermore, the fold-changes of the DAPs were signi cantly negatively correlated with their corresponding DEGs by Pearson's correlation tests (r = 0.03 and 0.11, p < 0.01, in PM_PS and PL_PM, respectively), indicating a poor correlation between transcript levels and protein abundance ( Fig. 6c-6d).

Identi cation of DAPs and DEGs associated with candidate pathways
To further clarify the biological functions of the co-regulated DEGs-DAPs genes, an enrichment analysis was conducted based on the GO analysis. The largest groups within the biological processes category were those linked with metabolic and cellular processes; catalytic activity and binding mainly included membranes, cell and cell parts were predominant in the category of molecular functions, both in PM_PS and PL_PM ( Fig. S4a-b).
Moreover, a pathway enrichment analysis was conducted in the PM_PS and PL_PM groups, based on the KEGG database. In the PM_PS, 14 DAPs were signi cantly enriched in 7 pathways, both in DEGs and DAPs, which including fructose and mannose metabolism pathway; phenylpropanoid biosynthesis; glutathione metabolism; ubiquinone and other terpenoid-quinone biosynthesis; pentose and glucuronate interconversions; amino sugar and nucleotide sugar metabolism, and galactose metabolism. In the PL_PM group, 4 DAPs were signi cantly enriched in the 3 pathways, both in the DEGs and DAPs, which included a calcium signaling pathway, pentose and glucuronate interconversions, and galactose metabolism pathway (Fig S4c-4d). The comparative analysis showed that 2 pathways, including pentose and glucuronate interconversions (2 DEGs and DAPs), and galactose metabolism pathways (2 DEGs and DAPs) were shared in the transcriptome and proteome data, for both the PM_PS and PL_PM groups. While the 2 DEGs and DAPs involved in the phenylpropanoid biosynthesis pathways were only shared by DAPs and DEGs in the PM_PS group. Therefore, involved in the shared DAPs and DEGs of the two pathways were further investigated as candidate genes related to the A. trifoliata fruit peel cracking.
Analysis of proteins expressed in A. trifoliata fruits identi es genes that might play relevant roles in fruit peel cracking Of the two candidate pathways, a total of 13 and 3 DAPs and 28 and 46 DEGs were detected in the PM_PS and PL_PM groups, respectively, where indicated that more DEGs than DAPs were involved in the peel cracking of A. trifoliata fruit ( Fig. 3a-3H). Moreover, most of these DEGs and DAPs were downregulated in the PM_PS group, and up-regulated in the PL_PM group, both in the transcriptome and proteome data. Notably, PL was up-regulated in the PM_PS group, and the PE and β-GAL2 genes were upregulated in the PL_PM both in the transcriptome and proteome data and showed strong positive correlations. While β-GAL1 had a negative correlation between the transcriptome and proteome data, as it was down-regulated in the transcriptome and up-regulated in the proteome. Furthermore, the results indicated that most of the genes encoding DAPs were not included in the DEGs, which was accorded with the observed differences between the proteome and transcriptome data.
Additionally, the signal transmissions underlying the A. trifoliata fruit peel cracking were studied by analyzing the protein-protein interactions using the STRING database. In the interaction network of these DEGs and DAPs (Fig. 7), there were two pathways identi ed, PE interacted with PL and βfructofuranosidase was interacted with ra nose synyhase. These results indicated that PE and PL should be further investigated as candidate proteins in the development of A. trifoliata fruit peel cracking.

Disscussion
The structural changes of the pericarp Fruit cracking is a key factor that in uences the utilization rate of biofuel feedstock, marketability of fruits, and causes signi cant loss of yield and commercial value. So elucidating the molecular mechanisms that regulate fruit cracking played key roles in the utilization of A. trifoliata for biofuels. But however there have been no studies and no identi ed genes in A. trifoliata that are related to fruit cracking. In this study, the structures of the different development stages of the fruits were observed, and there were signi cant differences in the different stages. In the non-cracking stage, the arrangement of pericarp cell and cuticle were dense, small intercellular space and distributed continuously. While the cells became thinner and the number of cell layers reduced in the initial cracking stages, and these characteristics increased at the total cracking fruit stage; the arrangement of the cells was loose and the cells began to degrade, compared with those not cracking (Fig. 1). These results were consistent with previous results that showed that the cell wall structures of the fruit pericarp had poor integrity, loose cell wall structures, deformed cell layers, presented larger spaces and break down cell wall during fruit cracking [25]. Studies have demonstrated that the arrangement of the subcutaneous layers of the cells were relatively regular, and that cell layers had a closer arrangement in the cracking-resistant tomato genotype [26]. The biomechanical behaviors of the plant cell walls changed due to the pectin degradation [27,28]. Moreover, the modi cations reduced the strength of the fruit pericarp, because the cell polysaccharides degraded the cell wall hydrolases, and the formation of phenolic cross-linking cell wall structural components catalyzed by cell wall peroxidase, resulted in changes to the pericarp development and fruit cracking [25]. The jujube fruit cracking might be related to the changes of cell wall structure, and the rearrangement of the cell wall at the later stages of fruit ripening [29]. Therefore, the changes of the structures and the ultrastructures of the pericarp cell wall between the unripe and ripe fruits may play a key role in the tendency of A. trifoliata fruits to crack.
The reference transcriptome and proteome generated by RNA-seq and TMT Fruit cracking is a complex phenomenon that is caused by a series of environmental, physiological, biochemical, and genetic changes during fruit ripening. In this study, the differences in the transcriptome and proteome were rstly investigated based on RNA-seq and TMT during different development stages.
Moreover, the transcriptome database was used for protein identi cation in this study, so the quality of the sequencing and assembly of the transcriptome data was crucial for the subsequent analyses.  [22], and 45.10%, 96.31% [1]. As there were no previous proteome studies reported on the Lardizabalaceae family, a total of 812 625 spectra, 68 151 identi ed spectra, 12 456 identi ed peptides, and 10 572 unique peptides, were found by proteomic analysis and 2839 proteins were identi ed in a reference proteome of the A. trifoliata pericarp. These results indicated that the A. trifoliata pericarp transcriptome and proteome presented here were comprehensive, accurate, and useful tools for future genetic research of A. trifoliata fruit cracking and of other Lardizabalaceae and fruit species.

Identi cation of potential regulators and metabolism pathways involved in fruit cracking
In this study, more DEGs were detected in the PL_PS group, while more DAPs were detected in the PM_PS group. The enrichment analysis of GO and KEGG pathway revealed that most of the DEGs and DAPs were involved in the metabolic process and ribosome metabolites. Moreover, more down-regulated and upregulated DEGs and DAPs were detected at the fruit cracking stage than the non-cracking stage. Previous studies indicated that numerous changes, such as increased respiration, fruit softening, metabolic activities in compounds, structural polysaccharides, and a softening of textures generated, accompanied the fruit ripening processes [32], suggesting that these changes may have some correlation with A.
The main components of plant cell walls, include lignin, cellulose, hemicellulose, and pectin, and these would be modi ed or degraded during cell differentiation and dehiscence in plants, and are linked to fruit cracking [32,33]. Previous studies have suggested that several cell-wall related genes are susceptible or resistant to fruit cracking, including CAD, 4CL, HCT; cell-wall modi cation genes-β-GAL, EXP, PE and PG; [34][35][36]; cellulase gene CELLULASE6 (CEL6) was also shown to be essential for silique dehiscence in Arabidopsis [37]. Additionally, cell-wall related transcription factor NAC also affected cell differentiation, seed abscission and fruit dehiscence [38]. In this study, cell wall-related DEGs and DAPs, including the galactose metabolism, phenylpropanoid pathway, pentose and glucuronate interconversions, starch and sucrose metabolism, amino sugar and nucleotide sugar metabolism showed consistent downregulation and upregulation in the cracking stage, compared with the non-cracking stage. Therefore, DEGs and DAPs, which are related to the cell wall synthesis and degradation was selected as candidates, and their gene expression levels were further investigated. Among them, 27 (13 in DEGs and 14 in DAPs) showed strongly correlated expression at the gene and protein expression levels, indicating that the cell-wall metabolic pathways may play a crucial role in the regulation of A. trifoliata fruit cracking, which is in accordance with the results of the structure changes generated in A. trifoliata pericarp.

DEGs and DAPs involved in candidate pathways determine A. trifoliata fruit cracking
In order to identify the candidate DEGs and DAP involved in the fruit cracking, an integrated quantitative proteomics and transcriptomics analysis was performed in this study. There were more DEGs (11205) than DAPs (240) in both the PM_PS and PL_PM groups, and 1904 shared DEGs and 17 shared DAPs were identi ed during the comparisons between the PM_PS and PL_PM groups. There was discordance between the transcript levels and protein abundance, which was similar to the reports for pomegranate [39] and rice [40]. The inconsistency between mRNA and proteins could be attributed to the posttranscriptional and post-translational regulation, reversible phosphorylation, splicing events in cells, and translation e ciency [39,41]. This inconsistency showed that the e ciency of gene translation and posttranslation processes is an important regulatory factor during A. trifoliata fruit development.
Therefore, the consistency between the transcriptional level and the trend of protein abundance between the transcript levels and protein abundance were performed using Pearson's correlation tests. In summary, a poor correlation (r = 0.03 and 0.11) between the protein abundance and the expression of the corresponding genes in the PM_PS and PL_PM groups. A similar result, showing a relatively poor correlation (r = 0.27-0.40) was identi ed in a previous study [42]. A possible explanation is that the regulation from mRNA to protein is a complex process, and the changes of the protein abundance were generated after its corresponding transcript was stabilizing [43]. Therefore, both the transcriptomic and proteomic data play key roles in deciphering the molecular processes involved in fruit cracking. and fruit-splitting processes [25,44]. For example, CanBGal3 displayed signi cant hydrolytic activity in cell wall pectin-degradation, and might play role in cell wall loosening [45]. Studies found that suppressing the PL gene could not only greatly increase the rmness of the full ripe fruits and reduce the postharvest softening in strawberry [46,47]; but also increased the cell separation, and the cellulose and hemicellulose contents, but reduced the water-soluble and total pectin in SlPL-RNAi fruits, compared with the WT, suggesting that the SlPL gene participated in the pericarp cell wall rearrangement during fruit softening [48]. Additionally, suppressing PE gene expression also signi cantly altered fruit cracking and viscosity, but had little effect on the fruit rmness. PE and β-Gals were differentially expressed in the cracked fruits, compared with the non-cracked fruits of the litchi pericarp, which were identi ed as candidate genes for fruit cracking in litchi [49]. Schuch et al. [50] indicated that suppressing PE gene expression signi cantly reduce cracking in tomato. Considered a key enzymes in lignin-degradation, PRX can induce cell wall loosening [51,52]. Consistent with these ndings, in this study, 6 proteins were functionally annotated among of thousands of identi ed DEGs and DAPs, both in transcriptome and proteome data. Among which, PE and PL gene were up-regulated both at gene and protein expression levels in both the PM_PS and PL_PM groups, during fruit ripening, and PE interacted with the PL protein by protein-protein interactions using the STRING database (Fig. 7 ). Moreover, D-Gal, PRX, and PRX2 were down-regulated in the PM_PS group and up-regulated in PL_PM group, both at gene and protein expression levels during fruit ripening. The signi cantly increased expression of these cell-wall related genes, especially the PE, PL and D-Gal genes, suggests their dynamic roles in A. trifoliata fruit cracking.
However, the BGal gene showed an opposite trend of expression at gene and protein expression levels in both the PM_PS and PL_PM groups, which may be because the changes in the protein abundance were generated after its corresponding transcript had been stabilized, so the role of the BGal gene in fruit cracking needs to be further studied. The regulatory functions of the other identi ed DEGs and DAPs in the pathways related to pentose and glucuronate interconversions, phenylpropanoid biosynthesis and galactose metabolism pathways have not been characterized, combined analysis of the other data may further illuminate their functions in relation to fruit cracking.

Conclusions
In conclusion, this study provides the rst comprehensive transcriptome and proteome data that indicate that there is a complex transcriptional and translational network in the regulating of fruit cracking in A.
trifoliate. These results suggested that these candidate genes/proteins may play important roles in fruit ripening and cracking of A. trifoliate fruits. The results described here provide important insights into A.
trifoliate fruit ripening and indicate that cell wall related genes/proteins play key roles in the process of fruit cracking and pave the way for further investigations into their molecular mechanisms and applications of A. trifoliate as a bioenergy crop.

Plant materials
The A. trifoliata sample Nong No.8 was used as the research material in this study. Nine year-old trees of Nong No.8 were growing in the nursery at Hunan Academy of agricultural sciences, Changsha, P. R.
China. According to our observations, in Hunan Province, the blooming stage for the germplasm of Nong No.8 was in early April, when 50% of the A. trifoliata owers were in bloom, and the fruit usually ripen in early October. Fruit of different developmental stages, including PS, PM and PL were randomly taken from the same Nong No.8 tree as a biological replicate when the fruits were ripe; in total, three biological replicates were collected for each stage. The pericarps of the fruits were cut longitudinally into several parts after harvesting and frozen in liquid nitrogen and stored at -80 ºC, prior to transcriptome, proteome, and qPCR analyses.
Anatomical structure of pericarp Anatomy of the pericarp samples taken from the Nong No.8 fruits at different development stages including PS, PM, and PL, were prepared for para n sections and scanning electron microscopy (SEM) was carried out according to Chen et al. [53]. Pericarp samples were xed directly in the eld using FAA [70% ethyl alcohol + 38% methyl aldehyde + 25% acetic acid (16:1:1)]. Then the tissues were subsequently dehydrated through an ethanol series with increasing ethanol concentrations and embedded in the para n. Subsequently, para n sections were stained with Safranin O-staining and observed by Axio Imager (Zeiss, Oberkochen, Germany), upright microscopy and images were displayed using Image-pro Plus 6.0 software.

RNA isolation, library construction and sequencing
Total RNA used for the RNA-seq assays was isolated from three independent replicates of pericarp in the PS, PM, and PL stages, as described by Tao et al [54]. The RNA samples were detected by absorbance ratio of A260/A280 with a Nanodrop ND-1000 system (Thermo Scienti c). Pair-end Libraries were prepared using NEBNext® UltraTM RNA Library Prep Kit for Illumina® (NEB, USA) following the manufacturer's instructions. The mRNA was puri ed from 3 µg of the total RNA using oligo (dT) magnetic beads followed by fragmentation carried out using divalent cations at elevated temperatures in NEBNext First Strand Synthesis Reaction Buffer. Subsequently, rst strand cDNAs were synthesized by random hexamer primer and Reverse Transcriptase (RNase H-) using mRNA fragments, as templates followed by the Second strand cDNA synthesis using DNA Polymerase I, RNAseH, buffer and dNTPs. The synthesized double-stranded cDNA fragments were then puri ed with AMPure XP system (Beckman Coulter, Beverly, USA). The puri ed double-stranded cDNA was polyadenylated and adapter-ligated for preparation of the paired-end library. Adaptor-ligated cDNA and adaptor primers were used for PCR ampli cation. PCR products were puri ed (AMPure XP system) and library quality was assessed on the Agilent Bioanalyzer 2100 system. Finally, sequencing was performed by Illumina HiSeq2500 instrument by Shanghai Applied Protein technology (Shanghai, China).

Quality control and transcriptome assembly
The raw paired-end reads of fastq format produced from the sequencing were rst processed with the in house Perl scripts. Those reads containing adapters, the excess "N" nucleotides with more than 10% of the bases, and reads of low-quality (reads with quality values ≤ 10) were removed by lter_fq software.
The Q20, Q30, GC-content, and sequence duplication levels of the obtained clean reads were calculated. The assembly of clean reads to unigene collections was performed using the Trinity software package (https://github.com/trinityrnaseq/trinityrnaseq/releases) [55]. The Trinity software consisted of three independent software modules, including Inchworm, Chrysalis, and Butter y, and the transcripts less than 200 bp in length were discarded. Sequences containing the longest cluster transcripts without redundancy extracted from transcripts can be considered unigenes.

Bioinformatics analyses
The de novo assembled unigenes were annotated in ve databases, which include NR, Pfam, the Swiss-Prot, GO, and KEGG pathway database, based on BLAST search with an E-value threshold of 1e − 5 .
Moreover, in order to further analyze the annotation results, GO and KEGG with an E-value of 1E − 5 were used for functional gene annotation. GO terms could be classi ed into three categories, including BP, MF and CC. In addition to the GO terms, the pathway maps were determined by KEGG database.
The normalized transcript abundance of the genes was estimated using the FPKM based on the length of the gene and reads count mapped to this gene. DESeq2 R package (1.16.1) software was used to identify the differential expression of the genes (DEGs), and the false discovery rate (FDR) was controlled using Benjamini and Hochberg's approach to adjusted P-value. Genes with an adjusted P-value < 0.05 and absolute fold change of 2 were deemed to be differentially expressed between the two samples. In addition, GO

Sequence database search and data analysis
The raw data were processed by MASCOT engine (Matrix Science, London, UK; version 2.2) and Proteome Discoverer 1.4 software was used to process MS/MS spectra. The search was performed using the following settings based on the A. trifoliata database: trypsin for the enzyme and 2 as the maximum missed cleavage allowed; xed modi cations of Carbamidomethyl (C), TMT-6plex (N-term) and TMT-6plex (K), variable modi cation of oxidation (M); mass tolerance for fragment ions of 0.1 Da while 20 ppm for peptide ions and both at peptide and protein levels of FDR less than 0.01, and only unique peptides of the protein were employed for the proteins identi cation and quanti cation. Proteins with a P value less than 0.05 and fold change ≥ 1.2 or ≤ 0.83 within a comparison were recognized as differentially abundant proteins (DAPs).
Functional categorization was performed using GO and KEGG pathways database with a P value ≤ 0.05. The protein functional network was performed by STRING 9.0 software (http://string-db.org). Clustering analysis of the DEPs was performed using Cluster 3.0 (http://bioservices.capitalbio.com/xzzq/rj/3885.shtml) and the Java Treeview software    Validation and expression analysis of selected genes and proteins using qPCR and RNA-seq. The expression levels of the genes revelaed by RNA-seq (Left y-axis) and qPCR (right y-axis). Histograms were gene expression detected by RNA-seq. Line graphs were relative expression validated by qPCR.

Figure 5
Validation and expression analysis of selected proteins and proteins using qPCR and TMT. The expression levels of the genes revelaed by TMT (Left y-axis) and qPCR (right y-axis). Histograms were protein abundance detected by TMT. Line graphs were relative expression validated by qRT-PCR.

Figure 6
Page 31/32 Correlations between mRNA and protein expression. a Venn diagram of genes quantified in the transcriptome (blue) and proteome (pink), DEGs(green) and DAPs (yellow) in PM_PS. b Venn diagram of genes quantified in the transcriptome (blue) and proteome (pink), DEGs(green) and DAPs (yellow) in PL_PM. c Scatterplot of the relationship betweengenes identified in both the transcriptome and proteome in PM_PS. d Scatterplot and correlation coefficients between DEGs and DEPs in PL_PM. e Scatterplot and correlation coefficients between DEGs and DEPs (the same trend) in PM_PS. F Scatterplot and correlation coefficients between DEGs and DEPs (the same trend) in PL_PM.

Figure 7
Analysis of the functional network by STRING 9.0 of the cell-wall related proteins, which are associated with pentose and glucuronate interconversions and galactose metabolism.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.