iTRAQ-Based Quantitative Glutelin Proteomic Analysis Reveals Differentially Expressed Proteins in Physiological Metabolism Process during Endosperm Development and their Impacts on Yield and Quality in Autotetraploid Rice

Autotetraploid rice, which is developed through chromosome set doubling using diploid rice, produces high-quality kernels that are rich in storage proteins. However, little information is available about the content of different proteins in autotetraploid rice and their proteomic analysis. changes in four storage proteins, albumin, globulin, and glutelin, analyzed in endosperm of autotetraploid at 15 DAF, while the peroxisome was richer in AJNT-4x at 20 DAF. The PPI network showed that ribosomal proteins gradually decreased with increasing endosperm development.


Background
Rice (Oryza sativa L.) is one of the most widely cultivated food crops in the world, especially in China, where the total output in 2019 was nearly 210 million tons, making China the world's largest rice producer. As the critical location of nutrient storage, endosperm in rice grains contributes greatly to embryo development for plant growth and nutrition, as rice is a staple food for humankind. For example, biological processes and protein production, aggregation, and metabolism are crucial for embryo development and nutritional quality.
The main ingredients of endosperm are starch and storage proteins. Endosperm storage proteins constitute approximately 5.5%-12% of polished rice grains (Zhang et al. 2014). With the development of modern biotechnology, mass spectrometry (MS)-based proteomics analysis has become a powerful tool to investigate protein expression levels in a broad range of biological samples, and it has also been used to interpret the molecular details of polyploid plants, such as wheat, peanut, and black locust (Islam et  As a basic feature in nearly all plant species, polyploidization normally results in a higher yield than that of diploid plants and is regarded as a powerful tool in crop breeding (Ramanna and Jacobsen 2003).
During the last decade, the application of polyploidization has been used in rice breeding, and several highly productive hybrid rice lines have been developed (Cai et al. 2007; He et al. 2011;; Ghaleb et al. 2020). Through comparative studies between diploid and autotetraploid rice cultivars, it was shown that there are higher glutelin contents in autotetraploid cultivars than in diploid cultivars (Cai 2011). Autotetraploid rice has a higher content of storage protein and better quality and nutritional value than those of its diploid counterpart.
Rice storage proteins, namely, albumin, globulin, prolamin, and glutelin, are classi ed into four fractions according to their differences in solubility. Glutelin and prolamin are the main storage proteins in endosperm. Glutelin constitutes approximately 60%-80% of total endosperm proteins and is the main focus in polished grains . Prolamin constitutes approximately 5%-10% of total endosperm proteins (Tanaka et al. 1980). Glutelin is easily digested ) and is rich in lysine (Ogawa et al. 1987). Therefore, glutelin has higher nutritional value than the other three endosperm proteins. Increasing the endosperm glutelin content will enhance the nutrient level of rice grains and increase the content of lysine, which is an essential amino acid for humans. This increase in glutelin content would be of profound signi cance for rice quality breeding in terms of cloning more novel glutelin genes and further studying their molecular mechanism.
Nevertheless, there is a lack of comparative investigations at the protein level between diploid and autotetraploid rice, which results in little understanding of the developmental process of endosperm after owering. In this study, a diploid and an autotetraploid rice line were researched further to identify differentially expressed proteins (DEPs) in different developmental stages. First, the protein components during endoplasm development and maturation were monitored, and some critical developmental time points were found. Then, a high-throughput isobaric tag for relative and absolute quantitation (iTRAQ) label-based proteomics analysis was performed to identify and quantify glutelin pro les in both diploid and autotetraploid endosperms, and bioinformatics tools were used to interpret the biological meaning of the changes in protein expression. The integration of these phenotypic and molecular results furthered the understanding of biological processes, primarily the regulation of protein expression and metabolism, in the development and maturation of rice endosperm. Furthermore, the differentially expressed genes will be further researched to elucidate the molecular mechanism of glutelin accumulation in autotetraploid rice.

Results
SDS-PAGE and phenotypic analysis of storage proteins in seed endosperm development The seed setting rate of new double autotetraploid rice is usually very low because of the separation behavior of the homologous chromosome. To study further dynamic changes in protein content and DEPs during different developmental stages, we selected AJNT-4x and AJNT-2x as the research materials. In our research, the seed setting rate of AJNT-4x reached 49.94%. On the other hand, the agronomic traits, such as grain length, grain width, and grain thickness, of AJNT-4x were also better than those of its diploid ( Table 1). The grain weight of AJNT-4x increased obviously, and there was a signi cant difference in grain weight between AJNT-4x and AJNT-2x. In addition, storage proteins in the seed endosperm were divided into albumin, globulin, prolamin, and glutelin fractions according to their solubility. SDS-PAGE analysis showed that our results were the same as those of a previous report. The molecular weights of the albumin and globulin fractions were 14-16 kDa and 26 kDa, respectively. The molecular weights of the prolamin fraction were 10 kDa, 13 kDa, and 16 kDa, but mainly 13 kDa. The molecular weights of the glutelin fraction were 57 kDa, 32 kDa, and 20 kDa; among these molecular weights, 57 kDa was a precursor protein, 32 kDa was an acidic subunit, and 20 kDa was a basic subunit. All the expression levels of the four seed proteins increased gradually with increasing owering time. The extracted protein fractions were divided into two parts: one was used to measure their contents, and the other was subjected to MS analysis for glutelin protein evaluation.
Dynamic changes in storage protein contents during endosperm development The contents and proportions of endosperm storage proteins affect the quality of rice (You et al. 2017). In this study, the contents of four storage proteins were measured by the BCA method (Noble and Bailey 2009). The four proteins all began to accumulate, and their contents were almost the same at 5 DAF in AJNT-2x and AJNT-4x.
The accumulation trends of albumin and globulin were almost the same in AJNT-4x and AJNT-2x. From 5 DAF to grain maturity, during the whole endosperm developmental period, the albumin and globulin contents increased only slightly ( Figure 1B).
The prolamin content rst decreased, followed by a slight increase, and then a rapid increase from 12 DAF. The content reached a high peak at 18 DAF in AJNT-4x, declined again, and nally stabilized.
Eighteen days after owering, slightly later than that in AJNT-2x, the prolamin content reached its maximum of 0.31 mg/mL in AJNT-4x. The prolamin content in AJNT-2x reached a peak of 0.18 mg/mL at 15 DAF. From 12 to 15 DAF, the prolamin content increased much more rapidly in AJNT-4x than in AJNT-2x ( Figure 1C). The prolamin content was much higher in AJNT-4x than in AJNT-2x during all the developmental stages, and the prolamin content difference between AJNT-4x and AJNT-2x at maturity was highly signi cant.
Compared with those of albumin, globulin, and prolamin, the variation range of glutelin was the largest during endosperm development. Glutelin accumulation was almost the same in AJNT-4x and AJNT-2x. There were two rapid accumulation stages: one was from 9 to 11 DAF, and the other was from 18 to 21 DAF in AJNT-4x, presenting two peaks at 11 DAF and 21 DAF ( Figure 1D). Then, the glutelin content declined gradually and stabilized. In general, the accumulation rate of glutelin in AJNT-4x was faster than that in AJNT-2x. The glutelin content was highly signi cantly different between AJNT-4x and AJNT-2x during the rapid accumulation period.
The change in the contents of the four proteins in AJNT-4x was almost the same as that in AJNT-2x, but the total content in AJNT-4x was much higher than that in AJNT-2x because the four proteins accumulated quickly in AJNT-4x ( Figure 1E). Glutelin is rich in lysine, which is an essential amino acid for humans, so we further studied changes in the glutelin content during the developmental stages of both AJNT-4x and AJNT-2x. According to the change in glutelin content during endosperm development, samples from several critical developmental stages, namely, samples collected 10, 15, and 20 DAF, were subjected to proteomic analysis by using the iTRAQ technique.
Proteome differential analysis at different endosperm developmental stages between AJNT-4x and AJNT-2x To investigate differential alterations in glutelin expression during seed endosperm development, glutelin fractions from both AJNT-4x and AJNT-2x endosperm were extracted following a TCA-acetone method (Méchin et al. 2007) and subjected to iTRAQ-based global proteomics analysis by LC-MS/MS (Unwin et al. 2010). A total of 841,375 spectra were generated, of which 38,851 were matched to known peptides from the utilized database. Eventually, 4,880 peptides, 4,047 unique peptides and 1,326 protein groups were identi ed. Comparative analysis between AJNT-4x and AJNT-2x at 10, 15, and 20 DAF was conducted, and 125, 159 and 78 DEPs were found (fold change>1.2 or<0.83, p-value<0.05) ( Table 2). There were many DEPs in the different developmental stages of rice endosperm analyzed. A total of 372 DEPs were measured at 10, 15, and 20 DAF AJNT-4x samples, which was much more than the 184 DEPs detected in the AJNT-2x samples.
Principal component analysis (PCA) showed that the autotetraploid rice endosperm had better separation than diploid endosperm, which indicated that more signi cant alterations in protein expression occurred in autotetraploid endosperm ( Figure 2).
Differentially expressed protein GO analysis for AJNT-4x and AJNT-2x All DEPs were annotated by GO. Comparing DEPs between AJNT-4x and AJNT-2x at 10, 15 and 20 DAF, it was surprising that no GO functions were mutual among the three comparison groups, but there were 15 common DEPs between 15 DAF and 20 DAF, and only one common DEP between 10 DAF and 15 DAF ( Figure 3A), which indicated that endosperm glutelin accumulation was regulated mainly by different DEPs during the late stage, and 15 DAF was a critical regulating point for glutelin accumulation.
The GO enrichment results also showed that the critical functions of DEPs exhibited little overlap at 10, 15 and 20 DAF. For instance, at 10 DAF, all DEPs between AJNT-4x and AJNT-2x endosperm samples were related to metabolic processes ( Figure 3B), including those of RNA (GO: 0016072, 0034660), peptides (GO: 0006518), proteins (GO: 0019538) and phosphate (GO: 0019220). At 15 DAF, the DEPs AJNT-4x and AJNT-2x represented a much broader range of regulatory functions, such as growth regulation (GO: 0040007). Metabolic processes or functions were still annotated at 15 DAF, but few proteins for RNA, peptide and protein metabolism were enriched ( Figure 3C). At 20 DAF, the number of DEPs between AJNT-4x and AJNT-2x endosperm was decreased compared to that at 10 or 15 DAF ( Figure   3D). However, the annotated functions of these proteins were still abundant and mainly enriched in stress response (GO: 0006950), carbohydrate metabolism (GO: 1901135), nutrient reservoir (GO: 0045735) and so on. In summary, differences in protein synthesis and metabolism mainly occurred at 10 DAF, but decreased gradually during later developmental stages. DEPs involved in cell growth and development and amino acid biosynthesis were mainly observed at 15 DAF.

KEGG analysis of differentially expressed proteinsbetween AJNT-4x and AJNT-2x
To investigate metabolic processes and feasible signaling pathways, these DEPs were subjected to KEGG analysis. The KEGG pathway results ( Figure 4) indicated that the biological pathways for AJNT-4x endosperm proteins were different from those for AJNT-2x endosperm proteins at 10, 15 and 20 DAF. Ribosomal proteins ( Figure S1), which were signi cantly higher in AJNT-4x than in AJNT-2x endosperm, were the major difference at 10 DAF due to alterations in central carbon metabolism and cellular senescence. At 15 DAF, the most enriched pathway was protein processing in the endoplasmic reticulum (ER) ( Figure S2), which contained 12 DEPs, and all of them were enriched in AJNT-4x endosperm compared to their expression levels in AJNT-2x. For example, protein disul de-isomerase and heat shock protein 70 are both critical for protein folding (Gruber et al. 2006; Wisén and Gestwicki 2008). The biosynthesis and metabolism of amino acids was another important series of pathways enriched at 15 DAF, and the enriched proteins were involved in arginine biosynthesis; alanine, aspartate and glutamate metabolism; tyrosine metabolism, cysteine and methionine metabolism; arginine and proline metabolism; phenylalanine metabolism; and phenylalanine, tyrosine and tryptophan biosynthesis. These amino acid biosynthetic and metabolic pathways were much higher in AJNT-4x than in AJNT-2x (Table S1). This result suggested that not only were protein synthesis and processing more active in AJNT-4x endosperm at 15 DAF but there was also a higher level of amino acid accumulation. Interestingly, two upregulated differential lysine proteins (B8AM24 and B8ARJ0) were found in the lysine biosynthesis process ( Figure  S3), and four upregulated differential proteins (P37866, Q9S768, A0A0N7KTS9 and B8AEL7) were found to regulate alanine, aspartic acid, and glutamic acid metabolism processes ( Figure S4).
In the late period (20 DAF) of endosperm development, the pathways of DEPs were primarily metabolic pathways, including fructose and mannose metabolism, methane metabolism, and starch and sucrose metabolism ( Figure S5). However, the most enriched pathway at 20 DAF was the peroxisome pathway ( Figure S6), in which 4 DEPs were downregulated in AJNT-4x when compared to their expression levels in AJNT-2x. As a crucial organelle that regulates large numbers of biological processes, for instance, metabolism and development (Hu et al., 2012), the lower level of the peroxisome pathway in AJNT-4x at this stage suggested reduced metabolic and developmental activity compared with that in AJNT-2x endosperm. Interestingly, KEGG pathway analysis also mapped DEPs to protein processing in the ER ( Figure S7) at 20 DAF, but two proteins, namely, protein transport protein Sec61 and transloconassociated protein, had lower expression levels in AJNT-4x than in AJNT-2x.

Protein-protein interaction (PPI) network for DEP analysis
The protein-protein interaction (PPI) network is another important tool to interpret proteomics information. DEPs between AJNT-4x and AJNT-2x were annotated, and PPI networks were generated by the String database (Szklarczyk et al. 2015). As shown in the results, most connected and annotated proteins were ribosomal proteins and metabolic proteins ( Figure 5). At 10 DAF, half of the DEPs in the PPI networks were ribosomal proteins, and the other half were metabolic proteins. With increasing endosperm development, ribosomal proteins gradually decreased, and metabolic proteins increased. Until 20 DAF, the proportion of ribosomal proteins was reduced to approximately 1/3, and no ribosomal DEPs were found in the PPI network.

WGCNA analysis of co-expression modules
Traditionally, the analysis of proteomics data has focused on DEPs, which were regarded as largely important participants in the comparison group. However, a large amount of "unchanged" proteins (fold change<1.2) and "nonsigni cantly changed" proteins (p-value>0.05) might be ignored because of disadvantages in statistics, although they might result in an important impact on the phenotype. Therefore, a WGCNA approach was applied to construct a global co-expression network without any arti cial cut-offs (Pei et al. 2017). First, the identi ed proteins were divided into 5 different modules ( Figure 6A), and then GO and KEGG analysis of each module were performed. Brie y, the blue module contained ribosomal and oxidative phosphorylation-related proteins; the brown module contained proteins related to carbohydrate metabolism, including the metabolism of amino/nucleotide sugars, starch, sucrose, fructose and mannose, and glycolysis; the turquoise module contained proteins related to the proteasome and fatty acid biosynthesis; the yellow and gray modules were similar to the blue module but failed to enrich any KEGG pathways ( Figure 6B). During the endosperm development process of AJNT-2x, it can be concluded that proteins in the blue and turquoise modules had high expression levels at 10 DAF that then decreased at 15 and 20 DAF. The proteins in the blue module were expressed much higher in AJNT-4x than in AJNT-2x at 10 DAF but had similar levels at 15 and 20 DAF ( Figure 6C). These ndings suggested that protein expression and oxidative phosphorylation, which provide energy to biological processes, were highly active in the early stage of development but quickly slowed. In contrast, carbohydrate metabolism processes were less active at 10 DAF than at 15 and 20 DAF but then increased in both AJNT-4x and AJNT-2x. Compared with the differences in expression levels between different developmental stages in AJNT-2x, those in AJNT-4x showed more signi cant differences.

Con rmation of proteomic data by qRT-PCR analysis
To validate the experimental proteomics data, thirteen randomly selected DEPs, including ten upregulated DEPs and three downregulated DEPs (fold change>1.2 or<0.83, p-value<0.05), were analyzed using realtime quantitative PCR, (Table 3). Among them, an ATP binding protein (A0A0N7KMN9) and a cell redox homeostasis protein (B8AJS5) were signi cantly downregulated, and a glutelin protein (Q40689) and cell growth protein (A3AI97) were upregulated, contrasting with the results of iTRAQ; the results for the other nine proteins were consistent with the iTRAQ data ( Figure 7). Thus, the proteomics data were reliable. Among the nine DEPs with qRT-PCR data correlating with those of iTRAQ, eight were upregulated proteins (Table 3). These proteins are very important to plant growth and development. For example, B8AM24 and B8ARJ0 are related to the lysine biosynthetic process, B8AQM6 regulates cell growth by extracellular stimulus, A2ZCE6 is related to endosperm development, and P55857 is a glutelin protein. These proteins will be subjected to functional experiments in subsequent work.

Discussion
The total storage protein content was higher in autotetraploid rice than in diploid Rice contains four seed storage proteins primarily located in the endosperm and is rich in all the eight essential amino acids necessary for human health (Kannan et al. 2001). In addition, the rice endosperm plays crucial roles in nourishing the embryo during embryogenesis and seed germination. Previous studies have shown that protein constructs were almost the same between autotetraploid rice and diploid rice, and the protein contents were higher in autotetraploid rice than in the corresponding diploid rice (Xie et al. 2007;Wang et al. 2008;Gu et al. 2015). In our study, the protein construction of four seed storage protein fractions was almost the same between AJNT-4x and its AJNT-2x, but the protein content was higher in AJNT-4x, which was also reported in previous studies. Through self-crossing of fty-two generations, the protein expression of the repetitive genome on homologous chromosomes tends to diploidize autotetraploid rice (Zhang et al. 2015), which presents similar protein construction. With increasing gene dosages, the protein expression content will be higher in autotetraploid rice than in the corresponding diploid rice. This is one of the causes of the increased seed setting rate of AJNT-4x.
The change in the protein contents was different between the four proteins in our study; the albumin and globulin content increased slowly, and the rate of increase was almost the same between AJNT-4x and AJNT-2x, differing only in the time at which the peak concentration was reached, i.e., their maximum concentrations were reached earlier in AJNT-4x than in AJNT-2x. The prolamin content increased more quickly in AJNT-4x than in AJNT-2x, but reached its maximum later in AJNT-4x than in AJNT-2x. The glutelin content began to increase from 6 DAF and reached its peak concentration at almost the same for AJNT-4x and AJNT-2x. These results differ from previous conclusions; for instance, prolamin was thought to be expressed beginning at 5 DAF, to be expressed abundantly at 10 DAF, and then to increase gradually until grains reached maturity (Juliano and Chemists 1985;Wang et al. 2009). This difference in results could be a consequence of different research materials. The beginning of endosperm cellularization was shown to be earlier in autotetraploids than in diploids (Wang et al. 2005), and protein accumulation may occur earlier in autotetraploids. The total storage protein content was higher in the autotetraploid than in the corresponding diploid rice in this study. In this study, we systematically analyzed these differentially expressed genes in the glutelin accumulation process at 10, 15, and 20 DAF between AJNT-4x and AJNT-2x by GO and KEGG pathway annotation, PPI networks, and WGCNA co-expression modules. The DEPs regulating protein synthesis and metabolism mainly occurred at 10 DAF. In this stage, the levels of ribosomal proteins in autotetraploid rice were signi cantly higher than those in diploid rice. The expression gradually decreased with endosperm development, while the DEPs involved in cell growth and acid synthesis, such as B8AQM6, A0A0N7KMN9, B8AJS5, and P37833, rapidly increased at 15 DAF. Finally, the DEPs promptly decreased from 105 to 36, with endosperm development from 15 DAF to 20 DAF. This result also showed that 15 DAF was the critical period of endosperm protein accumulation and that the DEPs were higher in the autotetraploid than in the diploid. Therefore, glutelin accumulated earlier and more in autotetraploid rice than in its diploid counterpart. The mass spectrometric results corresponded with the measured protein contents.
The amylose content was negatively correlated with protein content Although the proteomic pro les of endosperm development have rarely been reported, previous studies have been performed only at the transcriptome level (Zhou et al. 2013Zhu et al. 2019). Proteases synthesized by some of the genes identi ed in transcriptome studies were also identi ed in this study. For instance, granule-bound starch synthase I (waxy) (B1B5Z2), an important enzyme for amylose synthesis (Hanashiro et al. 2013), was identi ed and located in the gray WGCNA module, indicating that it was increasingly expressed in the late period of endosperm development. In the three different periods, the wax content of the autotetraploid was lower than that of the corresponding diploid, so the content of amylose synthase was also lower. The amylose content was negatively correlated with the protein content (Cheng et al. 2016), so the protein content of the autotetraploid was higher than that of the diploid.
Protein content was signi cantly negatively correlated with the number of grains per panicle and seed setting rate, and was positively correlated with grain length, grain width, and grain weight (Zou et al. 2001;Cheng et al. 2016). The number of grains per panicle and seed setting rate of the autotetraploid were all lower than those of the corresponding diploid, and the grain length, width and weight of the autotetraploid were also obviously higher than those of the corresponding diploid (Table 1).
However, some known genes involved in endosperm development, for example, KRP1 and some transcription factors, were unable to be identi ed in this dataset. This might be because of the acquisition of an incomplete proteomics dataset due to the data-dependent (DDA) acquisition method (Röst et al. 2014).

Changes in carbohydrate metabolism in the endosperm
Metabolism-related proteins were another main class of proteins that were identi ed in this study. As critical processes of energy metabolism, oxidative phosphorylation and glycolysis were highlighted in the proteomics results. Unlike oxidative phosphorylation, which was more active at 10 DAF, glycolysis-related proteins had relatively low expression levels at 10 DAF and showed different tendencies of increase in the diploid and autotetraploid endosperms (Figure 8). In the diploid endosperm, the expression of glycolysisrelated proteins continued to increase from 15 to 20 DAF, and in the autotetraploid endosperm, the peak of glycolysis occurred at 15 DAF and decreased slightly during the last 5 days. The alteration in carbohydrate metabolism changed not only the biological process but also the storage of nutrients.

Conclusion
In conclusion, a total of 372 DEPs were observed in AJNT-4x, and glutelin accumulation was regulated mainly by different DEPs during the late stage. We found that 15 DAF was a critical regulatory period for glutelin accumulation, and DEPs related to protein synthesis and metabolism mainly occurred at 10 DAF, while the differential proteins involved in cell growth and development and amino acid biosynthesis mainly occurred at 15 DAF. On the other hand, some important DEPs were found in the autotetraploid, and eight important upregulated proteins were veri ed by qRT-PCR, including B8AM24, B8ARJ0, B8AQM6, and A2ZCE6. These proteins regulate lysine biosynthesis, cell growth, and endosperm development. Our results revealed alterations of protein expression in these important biological processes during the different developmental stages in rice. These results will be helpful for gaining a general understanding of metabolic changes and molecular mechanisms related to autotetraploid rice development. The future direction of this study is to validate key proteins that are involved in developmental progress, and functional studies will also be performed on these key proteins, especially proteins that are still uncharacterized.

Materials And Methods
Grain weight and seed setting rate investigation and endosperm preparation Rice lines (O. sativa L. sp. Aijiaonante (AJNT)-2x and AJNT-4x after self-crossing for 52 generations) were grown on farms at South China Agricultural University in the early and late seasons in 2019. Ten rows were planted in each of the test plots, and the row spacing was 18 cm×21 cm. During the maturation period, a total of 70 rows were measured for seed setting in the middle of the test plots (Li et al. 2011). Calculations and analyses were performed using SPASS software ).

Protein extraction and quanti cation
Rice embryos were removed from the stored grains, and then the endosperms were frozen in liquid nitrogen and ground with a mortar and pestle. Five volumes of TCA/acetone (1:9, v/v) were added to the powder and mixed by vortexing. The mixture was placed at -20 °C for 4 h and centrifuged at 6,000 g for 40 min at 4 °C. The supernatant was discarded, and precooled acetone was added to wash the residue three times. The precipitate was then air dried.
For extraction of globulin, 100 mg of dried precipitate was resuspended in 1 mL of globulin extraction buffer (60 mM Tris-HCl, pH 6.6, 0.7 M NaCl), mixed and placed in a 37 °C water bath for 20 min and then centrifuged at 11,000 rpm for 10 min. The supernatant was collected as the globulin fraction. The precipitate from the last step was washed with water, resuspended in 65% propanol and then incubated in a 37 °C water bath for 20 min. The supernatant was collected after centrifugation at 11,000 rpm for 10 min as the prolamin fraction. Finally, the glutelin fraction was extracted by a 1.2% lactic acid solution using the same method described above. To prepare samples for analysis via liquid chromatography-tandem mass spectrometry (LC-MS/MS), 30 volumes of SDT buffer were added to 20-30 mg of powder, mixed and boiled for 5 min. The lysate was sonicated and then boiled for 15 min. After centrifugation at 14,000 g for 40 min, the supernatant was ltered through a 0.22 µm lter. The ltrate was quanti ed with the BCA protein assay kit (Bio-Rad, USA). All samples were stored at -80 °C.

Protein digestion and iTRAQ labeling
Protein digestion was performed according to the FASP procedure described by Wiśniewski et al. (2009), and the resulting peptide mixture was labeled using the 8-plex iTRAQ reagent according to the manufacturer's instructions (AB Sciex). Brie y, 200 μg of protein for each sample was incorporated into 30 μL of STD buffer (4% SDS, 100 mM DTT, and 150 mM Tris-HCl; pH 8.0). The detergent, DTT, and other low-molecular-weight components were removed using SDT buffer (4% (w/v) SDS in 100 mM Tris-HCl; pH 7.6; 50 mM DTT) by repeated ultra ltration (Microcon units, 30 kD). Then, 100 μL of 0.05 M iodoacetamide in UA buffer was added to block reduced cysteine residues, and the samples were incubated for 20 min in the dark. The lters were successively washed with 100 μL of SDT buffer three times and 100 μL of DS buffer (50 mM triethylammonium-bicarbonate at pH 8.5) twice. Finally, the protein suspensions were digested with 2 μg of trypsin (Promega) in 40 μL of DS buffer overnight at 37°C , and the resulting peptides were collected in the ltrate. The peptide content was estimated by UV light spectral density at 280 nm using an extinction coe cient of 1.1 for a 0.1% (g/L) solution that was calculated on the basis of the frequency of tryptophan and tyrosine in vertebrate proteins.
For labeling, each iTRAQ reagent was dissolved in 70 μL of ethanol, added to the respective peptide mixture, and multiplexed and vacuum dried.

LC-MS/MS analysis
Experiments were performed on a Q Exactive mass spectrometer that was coupled to an Easy nLC system (Proxeon Biosystems, now Thermo Fisher Scienti c). Ten microliters of each fraction was injected for nanoLC-MS/MS analysis. The peptide mixture (5 μg) was loaded onto a C18-reversed-phase column (Thermo Scienti c Easy Column, 10 cm long, 75 μm inner diameter, 3 μm particle size) in buffer A (0.1% formic acid) and separated with a linear gradient of buffer B (80% acetonitrile and 0.1% formic acid) at a ow rate of 250 nL/min controlled by IntelliFlow technology over 140 min. MS data were acquired using a data-dependent top10 method that dynamically chose the most abundant precursor ions from the survey scan (300-1,800 m/z) for high-energy collisional dissociation (HCD) fragmentation. Determination of the target value was based on predictive automatic gain control (pAGC). The dynamic exclusion duration was 60 s. Survey scans were acquired at a resolution of 70,000 at m/z 200, and the resolution for HCD spectra was set to 17,500 at m/z 200. The normalized collision energy was 30 eV, and the under ll ratio, which speci es the minimum percentage of the target value likely to be reached at the maximum ll time, was de ned as 0.1%. The instrument was run with peptide recognition mode enabled. Protein interaction data of the studied proteins were retrieved from the IntAct molecular interaction database6 by their gene symbols. The results were downloaded in XGMML format and imported into Cytoscape 3.7.0 for further analysis. A weighted gene co-expression network analysis (WGCNA) coexpression network was generated by the R package (Langfelder and Horvath, 2008).

Quantitative PCR
Quantitative PCR was conducted using the manufacturer instructions of the 2×miRNA qPCR masterMix kit of Sangon Biotech (Shanghai) Co., Ltd. A 0.2 mL thin-walled 96-well PCR plate was acquired and numbered. The 2×miRNA qPCR masterMix dye (10 μL) was added to a tube; then, 1 μL of each of the forward and reverse primers (primer concentration: 10 μmol/L) was added, and 1 μL of the mixed cDNA was added to the tube. Each tube was supplemented with ddH 2 O to 20 µL, centrifuged at 800 rpm for 30 s, and placed in a LightCycler 480 Real-time PCR System. A two-step cycle was carried out: 95 °C for 30 s, 95 °C for 5 s, 60 °C for 1 min, and the last two processes for 40 cycles and 40 °C for 1 min. The 2×miRNA qPCR masterMix kit was purchased from Sangon Biotech (Shanghai) Co., Ltd.

Availability of Data and Materials
The data sets supporting the conclusions of this article are included within the article and its additional les.

Ethics Approval and Consent to Participate
Not applicable.

Consent for Publication
Not applicable.

Competing Interests
The authors declare that they have no competing interests   Table 3 Real-time quantitative PCR analysis for thirteen DEPs. Figure 1 Glycolysis pathway. Each circle represents a metabolite, and proteins have been identi ed in blue according to the associated heatmap located on the right. In the heatmap, the top row, labeled Diploid, represents AJNT-2x, and the bottom row, labeled Tetraploid, represents AJNT-4x. Each column represents a trait. The heatmap is color-coded by correlation according to the color legend.       Endosperm protein content analysis of different developmental stages between Aijiaonante (AJNT)-4x

Figures
and AJNT-2x A, standard curve; B, albumin and globulin content variation; C, prolamin content variation; D, glutelin content variation; E, total storage protein content at maturity. "*" indicates a signi cant difference at the P=0.05 level; "**" indicates a highly signi cant difference at the P=0.01 level.

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