Characterization of seed germination in LS and LR
The maximum germination (gMAX) of LR was about 52% while that of LS reached 94% (Fig 1a), consistent with the previous study results [17], suggesting that the low germination character was steadily inherited in LR. According to the germination curve, the rapid growth period was from 3d to 8d after incubated (DAI). Seed germination is a process with complex physiological, biochemical and molecular biological basis [28]. With the absorption of water, the storage macromolecules such as starch, proteins and lipids were decomposed into small and soluble molecules that can be easily utilized and transformed to support the germination [29]. The content of soluble sugar in the 0 DAI seeds of LS was higher than LR. After incubation, the soluble sugar content deceased and then dramatically increased at 6 DAI in LS and LR seeds. The soluble sugar content of LR seeds was always lower than that of LS seeds until 8 DAI (Fig. 1b). Similarly, the content of soluble protein showed down first and then up, and the content in LR was lower than that in LS during the germination process (Fig. 1c). These results indicated that more soluble sugar and protein could directly participate carbon oxidation and enzyme catalysis to supply energy in LS in comparison with LR during germination. Combined with the dynamic change of germination rate and soluble sugar and protein content, we chose 3 DAI and 6 DAI seeds for transcriptome analysis to compare the difference of germination process between LS and LR.
Hormones regulate the seed germination of LS and LR
Gibberellin (GA) and abscisic acid (ABA) antagonistically regulate the transition of germination and dormancy, that GA promote germination and ABA promote dormancy [30] To investigate whether the germination delay of LR was related to the hormonal regulation, the seeds of LS and LR were treated with exogenous GA and ABA and their synthesis inhibitors, fluridone (FL) and paclobutrazol (PA), to observed the gMAX values (Fig. 1d). The application of GA significantly evaluated the gMAX of LR from 52% to 78%, suggesting that exogenous GA could promote the germination of LR. The application of ABA had no significant effect on the gMAX of LS but decrease that of LR to 17%, suggesting that the inhibitory by ABA was stronger in LR than LS. PA and FL inhibits the GA and ABA biosynthesis had been proved in many plant species [31]. The application of PA decreased the gMAX of LS to 73% and that of LR decreased to 25%, suggesting that the GA biosynthesis in LR was more strongly inhibited by PA. The application of FL dramatically increased the gMAX of LR to 94%, suggesting that the dormancy of LR completely removed when ABA biosynthesis was inhibited.
de novo assembly of P.fugx reference transcriptome
Two P. fugax accessions LS and LR were incubated for 3d and 6d, respectively, and three biological replicates were designed in each accession at two time points. Thus twelve samples were used to conduct the RNA-seq analysis. The number of total clean reads of each samples were shown in Table 1. A total of 476,220 transcripts and 149,330 genes were de novo assembled, and the distribution of length interval was shown in Fig. 2a.
The number of annotated unigenes of each database was show in Table 2. Basically, a total of 99,209 genes (66.43%) were at least annotated to one database, suggesting that the de novo transcriptome had relatively complete gene function information. Among them, 67,183 genes (44.98%) were annotated to NR database, and established that P.fugax were highly similar to Aegilops tauschii subsp. tauschii (30.4%), Brachypodium distachyon (L.) P. Beauv. (16.5%), Hordeum vulgare subsp. vulgare (8.8%), Triticum urartu (5.6%), and Oryza sativa Japonica Group (4.4%) (Fig. 2b). A total of 20,042 genes (13.42%) were annotated into 25 classifications of KOG database (Additional file 2). Among them, the highest number of categories were “posttranslational modification, protein turnover, chaperones” (2,701, 13.48%), “general function prediction only” (2,627, 13.11%) and “translation, ribosomal structure and biogenesis” (2,321, 11.58%) (Fig. 2c). Moreover, there were 61,384 genes (41.1%) annotated to GO database, 22,834 unigenes (15.29%) annotated to KEGG database, 68,540 (45.89%) to Nt database, and 52,642 (35.25%) to SwissProt database (Additional file 3, 4).
Identification of differentially expressed genes (DEGs)
To determine the molecular basis involved in the germination process of LS and LR, two comparison settings, LS_6d vs LS_3d and LR_6d vs LR_3d, were analyzed. The biological samples had good reproducibility due to the high correlation values (Fig. 3c). The distribution of the DEGs in LS and LR were directly exhibited in volcano plots (Fig. 3a and 3b). A total of 11,856 DEGs were identified and 8,936 were up-regulated and 2,920 were down-regulated in the LS comparison setting. A total of 23,123 DEGs were identified and 20,055 were up-regulated and 3,068 were down-regulated in the LR comparison setting. A venn diagram showed that 7,165 up-regulated DEGs and 1,167 down-regulated DEGs were common in LS and LR comparison settings (Fig. 3d).
GO and KEGG enrichment of DEGs
To understand the difference in biological function and processes related to germination between LS and LR from 3 DAI to 6 DAI, all the DEGs were enriched to GO and KEGG database. The up-regulated DEGs in LS were enriched to 109 GO terms and the down-regulated DEGs were enriched to 9 GO terms (Additional file 5). The up-regulated DEGs in LR were enriched to 224 GO terms and the down-regulated DEGs were enriched to 40 GO terms (Additional file 6). The up-regulated DEGs in LS and LR were highly enriched in “metabolic process”, “single-organism process”, and “catalytic activity”. With regard to the down-regulated DEGs, the enriched GO terms were much less in LS than LR. These results indicated that most of the up-regulated DEGs in both LS and LR had similar biological function. All the up-regulated DEGs in LS and LR were enriched to KEGG database and the top 20 enriched pathways were shown in Fig. 4a and 4b. The majority of up-regulated DEGs were related to lipid metabolism, carbohydrate metabolism, amino acid metabolism and secondary metabolites biosynthesis in LS and LR seed during germination. Here, we selected some pathways to investigate the difference of regulation network in LS and LR (Table 3).
DEGs related to hormones biosynthesis and signal transduction
Seeds germination and dormancy processes are regulated by diverse endogenous hormones. According to the results of exogenous hormones regulation, GA and ABA affected the seed germination of LS and LR at different levels. The genes (Cluster-37472.24032 and Cluster-37472.23672) coding two ent-kaurene synthases (KS) and gene (Cluster-37472.70387) encoding gibberellin 3beta-hydroxylase (GA3ox) were up-regulated in both LS and LR comparison settings, and the increasing was much higher in LS than that in LR. The first step in GA biosynthetic pathway is transformed geranylgeranyl pyrophosphate to ent-kaurene catalyzed by ent-kaurene synthetases. The KS deficient mutants of Arabidopsis showed strong seed dormancy and recovered germinate with exogenous GA treatment [32, 33]. GA3ox catalyzes the final biosynthetic step to produce bioactive GAs. Two Arabidopsis genes, GA4 and GA4H, encoding GA3ox were expressed during seed germination [34]. These GA biosynthesis related genes with higher expression in LS suggested that GA biosynthesis was more active in LS than LR. Consistent with this result, the inhibition of GA biosynthesis by PA was more effective in LR than LS. The gene (Cluster-51396.0) encoding phytochrome-interacting factor 4 (PIF4) was down-regulated in LS but up-regulated in LR during germination. PIF4 regulated the gibberellin-signaling pathway via DELLA proteins, which blocked the GA signal pathway and resulting in germination delay [35]. Thus, the increased expression of PIF4 might repress seed germination via inhibiting the GA signal transduction in LR.
ABA is the major hormone that inducing seed dormancy. The gene (Cluster-37472.23110) coding 9-cis-epoxycarotenoid dioxygenase (NCED) was down-regulated in LS but up-regulated in LR during germination. The expression of gene (Cluster-37472.23992) coding phytoene synthase (PSY) and gene (Cluster-37472.76697) coding lycopene epsilon-cyclase (LcyE) were both elevated in LS and LR during germination, and the increasing was higher in LR than LS. The cleavage of 9-cis-epoxycarotenoids catalyzed by NCED is the key regulatory step of ABA biosynthesis. The Arabidopsis mutants of NCED genes, Atnced6 and Atnced9, reduced ABA content level in seeds [36]. PSY is the limiting step of carotenoids synthesis, the upstream of ABA biosynthesis. The PSY gene overexpressed Arabidopsis mutant exhibited delayed germination, and the degree of delay was positively associated with the increased levels of carotenoids and ABA [37]. The gene (Cluster-37472.6593) encoding serine/threonine-protein kinase SRK2 (SNRK2) and gene (Cluster-48355.0) coding ABA responsive element binding factor (ABF) were up-regulated in LS and LR, and increasing were higher in LR than LS. It has established in rice and Arabidopsis that SNRK2 is activated by ABA and phosphorylate ABF, which is important for the activation of ABA signal transduction [38, 39]. Taken together, these genes associated with ABA biosynthesis and signal transduction pathways exhibited higher expression level in LR than LS, which might explain why the LR seeds were more sensitive to ABA and FL could restore the germination ability of LR.
DEGs related to fatty acid metabolism
ACCase catalyzes the carboxylation of acetyl-CoA to malonyl-CoA, acting as the initial step of fatty acid biosynthesis [40, 41]. ACCase has three subunits: biotin carboxylase carrier protein, biotin carboxylase, and carboxyltransferase domains [42]. Molecular and biochemical studies have proved that the carboxyltransferase domain is the target site of ACCase-inhibiting herbicides [43]. LR plants had an amino acid substitution in carboxyltransferase domain that caused the resistance to ACCase-inhibiting herbicide [16]. Unfortunately, the transcriptome data did not identify any DEGs annotated as ACCase carboxyltransferase domain in LS or LR. But gene (Cluster-39490.0) coding ACCase biotin carboxylase (ACACA) was identified to up-regulated with the higher increasing level in LS than LR. ACCase is the rate-limiting enzyme in fatty acid metabolism, the different regulation level between LS and LR might influence the downstream reactions as well. The gene (Cluster-31226.0) coding 3-oxoacyl-[acyl-carrier protein] reductase (FabG), gene (Cluster-37472.4287) coding long-chain acyl-CoA synthetase (ACSL) showed higher regulation level in LS compared with LR during germination. These enzymes were respectively involved in the elongation cycle and termination of fatty acid synthesis. Besides, the genes involved in fatty acid degradation were also up-regulated higher in LS than LR, such as the gene (Cluster-33548.0) coding acetyl-CoA acyltransferase 1 (ACAA1), gene (Cluster-37472.85335) coding acyl-CoA dehydrogenase (ACADM), gene (Cluster-30044.0) coding acetyl-CoA C-acetyltransferase (ACAT). The fatty acid forms triacylglycerol (TAG) that degrades to provide carbon and energy during germination and early seedling growth. Some ACCase herbicide resistant weeds also showed higher level of dormancy, absence of germination in dark conditions, and delayed seed germination [44-46]. We assumed that the mutated ACCase might alter normal fatty acid metabolism and impact the germination.
DEGs related to carbohydrate metabolism
Acetyl-CoA, acts as the precursor of fatty acid biosynthesis, is also the product of glycolysis and TCA. According to the dynamic change of soluble sugar and protein content, more soluble sugar and protein could directly participate carbon oxidation and enzyme catalysis in LS than LR during germination. Thus, the genes involved in carbohydrate metabolism were selected to compare the differential expression pattern. The gene (Cluster-37472.6545) coding glucose-6-phosphate isomerase (GPI),gene (Cluster-37472.71503) coding fructose-bisphosphate aldolase (FBA), gene (Cluster-35171.2) coding phosphoenolpyruvate carboxykinase (PCKA), gene (Cluster-40718.0 ) coding pyruvate dehydrogenase E1 component alpha subunit (PDHA), genes (Cluster-37472.39187, Cluster-37472.47005) coding alcohol dehydrogenase (ADH), were up-regulated in LS and LR comparison settings, and the increasing levels were higher in LS than LR. All of these genes were involved in glycolysis, which produces energy by utilizing endosaccharides [47]. PCKA catabolizes the storage lipid and protein to produce soluble sugar via glycolysis. The PCKA deficient mutants of Arabidopsis and tomato were observed growth suppression of germinated seedlings [48].
The genes (Cluster-37472.85108, Cluster-15133.3) coding malate dehydrogenase (MDH), gene (Cluster-37472.21027) coding citrate synthase (CS), gene (Cluster-33485.0) coding 2-oxoglutarate dehydrogenase E1 component (OGDH), genes (Cluster-38121.0, Cluster-34010.0) coding succinyl-CoA synthetase alpha subunit and beta subunit (LSC1 and LSC2), genes (Cluster-37472.82495, Cluster-39676.0) coding succinate dehydrogenase (ubiquinone) iron-sulfur subunit (SDHB), and gene (Cluster-41938.0) coding pyruvate carboxylase (PYC) were up-regulated in LS and LR at 6 DAI compared with 3 DAI, and the increasing levels were higher in LS than LR. All of these genes were involved in TCA cycle, which produced high-energy phosphate compounds as the main source of cellular energy [49]. MDH catalyze the interconversion of malate and oxaloacetate in TCA cycle. The Arabidopsis MDH gene mutants showed slowly germination rate, higher content of free amino acids, different sugar levels, and lower content of 2-oxoglutarate [50]. The seeds of the CS deficient Arabidopsis were dormant and did not metabolize triacyglycerol [51].
The gene (Cluster-37472.72264) coding 6-phosphogluconate dehydrogenase (PGD), gene (Cluster-37472.83492) coding transketolase (TKT), gene (Cluster-37472.237) coding gluconokinase (gntK) were highly up-regulated in LS compared with LR during germination. Pentose phosphate pathway takes substrate from glycolysis and feeds its products back into glycolysis, which is important in determining the flux through glycolysis [52].
One of the major changes during germination is a rapid increase in respiration, which involves glycolysis, oxidative pentose phosphate pathway, TCA cycle and oxidative phosphorylation. All of the DEGs mentioned above were related to TCA cycle, glycolysis or pentose phosphate pathway that producing energy during seed germination. These genes were up-regulated in LS and LR during germination and had higher expression in LS, indicating that more energy were supplied to promote germination in LS in comparison with LR.
Validation of DEGs by qRT-PCR
A total of sixteen DEGs were randomly selected to verify the accuracy and reproducibility of the transcriptome results by qRT-PCR (Fig. 5). The correlation between transcriptome results (FC) and qRT-PCR results (2-ΔΔCt) were calculated using log2 fold variation measurements to produce a scatter plot. The results showed that the expression profiles of these DEGs were consistent with the transcriptome results, with relative R2 = 0.8751 and 0.7376 in LS and LR, respectively.