Comprehensive analysis of transcriptional regulation and functional genes induced by alkaline salt stress in roots of two switchgrass (Panicum virgatum L.) genotypes

Background: Soil salinization is a major limiting factor for crop cultivation. Switchgrass is a perennial rhizomatous bunchgrass that is considered an ideal plant for marginal lands, including sites with saline soil. Here, we investigated the physiological responses and transcriptome changes in the roots of two switchgrass genotypes under alkaline salt stress. Results: Alkaline salt stress signicantly affected the membrane, osmotic adjustment and antioxidant systems in switchgrass roots, and the ASTTI values between Alamo and AM-314/MS-155 were divergent at different time points. A total of 108,319 unigenes were obtained after reassembly, including 73,636 unigenes in AM-314/MS-155 and 65,492 unigenes in Alamo. A total of 10,219 DEGs were identied, and the number of upregulated genes in Alamo was much greater than that in AM-314/MS-155 in both the early and late stages of alkaline salt stress. The DEGs in AM-314/MS-155 were mainly concentrated in the early stage, while Alamo showed greater advantages in the late stage. These DEGs were mainly enriched in plant-pathogen interactions, ubiquitin-mediated proteolysis and glycolysis/gluconeogenesis pathways. We characterized 1,480 TF genes into 64 TF families, and the most abundant TF family was the C2H2 family, followed by the bZIP and bHLH families. A total of 1,718 PKs were predicted, including CaMK, CDPK, MAPK and RLK. WGCNA revealed that the DEGs in the blue, brown, dark magenta and light steel blue 1 modules were associated with the physiological changes in roots of switchgrass under alkaline salt stress. The consistency between the qRT-PCR and RNA-Seq results conrmed the reliability of the RNA-seq sequencing data. A molecular regulatory network of the switchgrass response to alkaline salt stress was preliminarily constructed on the basis of transcriptional regulation and functional genes. Conclusions: The alkaline salt tolerance of switchgrass may be achieved by the regulation of ion homeostasis, transport proteins, detoxication, heat

When plants are exposed to alkaline salt stress, the root is the rst tissue affected, and various biochemical and physiological mechanisms are stimulated to deal with the stress. Alkaline stress greatly reduced root growth and root vigour and induced a marked accumulation of superoxide anions (O 2 .-) and H 2 O 2 in rice roots [11]. The increasing pH around the root system also leads to the deposition of metal ions, resulting in the reduction of inorganic anions and the hindrance of plant uptake of mineral nutrients [12]. The roots minimize the distribution of absorbed salt at the tissue and cellular levels to avoid accumulation of toxic concentrations in the cytosol of functional leaves [13]. Wheat roots exhibited greater growth performance in response to alkaline salt stress as a result of increased glutamine synthetase activity and soluble protein contents [14].
Plants respond to alkaline salt stress by regulating the expression of a variety of salt-responsive genes, such as osmoregulatory genes [15], transporters/antiporters [16], transcription factors (TFs) [17], and protein kinases (PKs) [18,19]. Genes encoding Na + transport proteins are involved in regulating Na + transport under alkaline salt stress [20]. A quantitative trait locus (QTL) detected from NaCl and NaCO 3 treated rice suggested that the genes controlling the transport of Na + , in the form of NaCl and NaHCO 3 , may be different or induced in an uncoordinated manner by salt stress [21]. Vacuolar proton pump ATPase (V-H + -ATPase) is a multisubunit membrane protein complex that plays a major role in the activation of ion and nutrient transport. Overexpression of ScVHA-B, ScVHA-C and ScVHA-H improves tolerance to alkaline salt stress in transgenic alfalfa [22]. The HD-Zip TF family is one of the largest plantspeci c TF superfamilies and plays important roles in the response to abiotic stresses. Gshdz4, an HD-Zip gene, showed a high response to alkaline stress in wild soybean treated with 50 mM NaHCO 3 [23].
MsCBL4 plays an important role in alkaline salt stress tolerance via its in uence on the regulation of calcium transport and accumulation [24]. Overexpression of GsMSRB5a and GsCBRLK in Arabidopsis enhanced alkaline stress tolerance, inhibiting ROS accumulation and modifying the expression of ROS signalling, biosynthesis and scavenging genes. With the continuous development of high-throughput sequencing, some genes and TFs have been identi ed in switchgrass [25,26]. These interacting genes form multiple pathways, such as the salt overly sensitive (SOS) pathway, the calcium-dependent protein kinase (CDPK) pathway and the mitogen-activated protein kinase (MAPK) pathway [27].
Switchgrass (Panicum virgatum L.), a perennial warm-season C 4 rhizomatous bunchgrass, shows great prospects in terms of its strong adaptability, high water and nitrogen use e ciency, rapid growth and high productivity [28]. As an important ethanol bioenergy plant, switchgrass is considered one of the ideal plants to alleviate soil salinization and can still grow well in alkaline saline soil and marginal soil [29].
Anderson et al. showed that three lowland type switchgrass varieties (EG 2101, EG 1101 and EG 1102) had higher emergence rates and biomass yields under moderate to severe salt stress [28]. However, the molecular mechanism of switchgrass tolerance to alkaline salt stress is not well understood. There is much potential to be exploited in the alkaline tolerance of switchgrass.
To understand the responses and molecular mechanisms induced by alkaline salt stress in switchgrass, we analysed the relevant transcriptional regulation and functional genes by assaying the physiological and transcriptome changes in roots of two switchgrass genotypes (Alamo and AM-314/MS-155) under alkaline salt stress for a period of 24 h. The pattern of association between differentially expressed genes (DEGs) and physiological changes in response to alkaline salt stress were explored by weighted gene coexpression network analysis (WGCNA). According to the expression patterns of signal transduction, kinases and binding proteins, TFs, and physiological responses, a regulatory network model based on functional genes was established to comprehensively illustrate the alkaline salt stress tolerance mechanism of switchgrass.

Results
Changes in ASTTI in two switchgrass genotypes The alkaline salt tolerance trait index (ASTTI) was used to evaluate the physiological responses of Alamo (alkali-tolerant genotype) and AM-314/MS-155 (alkali-sensitive genotype) under alkaline salt stress for 0, 3, 6, 12 and 24 h. Ten physiological traits, including the relative water content (RWC), relative electrical conductivity (REC), the contents of malondialdehyde (MDA), free proline, soluble protein, soluble sugar, and reduced glutathione (GSH), and the activities of superoxide dismutase (SOD), peroxidase (POD) and catalase (CAT), were divided into three groups for subsequent analysis. The RWC, REC and MDA contents were used as indicators of the membrane system, and the ASTTI value of the root RWC (Fig. 1A) showed an upward trend, while the ASTTI values of the REC and MDA content showed a downward trend ( Fig. 1B and C). The root REC and MDA contents in AM-314/MS-155 were signi cantly higher than those in Alamo at 4 h post alkaline salt stress. The contents of free proline, soluble protein and soluble sugar were used as indicators of the osmotic adjustment system, and the osmotic regulation level of the two cultivars rst increased with increasing stress duration and then decreased ( Fig. 1 D-F). The osmotic regulation ability of Alamo was signi cantly higher than that of AM-314/MS-155 and was highest at 0, 6 and 24 h. For the indicators of the antioxidant system (GSH, SOD, POD, CAT), the activities of SOD, POD and CAT of Alamo were signi cantly higher than those of AM-314/MS-155 ( Fig. 1H-J), while the content of GSH, a nonenzymatic substance, was signi cantly lower than that of AM-314/MS-155 at the early stage ( Fig.  1G). In addition, we observed that there were signi cant physiological differences between the two varieties at different time points under alkali stress, especially at 6 and 24 h. Therefore, we chose these two time points for further transcriptome analysis.
De novo assembly and annotation of unigenes Six cDNA libraries were constructed from total RNA extracted from E5 stage roots of Alamo and AM-314/MS-155 treated with alkaline salt stress for 0, 6 and 24 h. A total of 114.03 Gb of clean data were obtained, with 6.05 Gb of clean data for each sample and a Q30 base percentage of 93.21% or more (Additional le 1: Table S1). Since the whole-genome sequence of switchgrass is not currently publicly available, valid readings from the six libraries were combined for reassembly (Additional le 2:  Figure S1A). The gene expression levels in response to alkaline salt stress were evaluated by converting the mapped read count for each gene into the expected number of fragments per kilobase of reproduction per million mapped reads (FPKM) to eliminate the effects of transcript length and sequencing differences on computational expression (Additional le 5: Table S4). The gene expression levels were not evenly distributed in the different stress environments in the boxplot diagram of the FPKM values (Additional le 4: Figure S1B). Then, the correlations of each biological sample were evaluated by Pearson correlation coe cients, and a value of r 2 close to 1 indicated a strong correlation between two replicate samples (Additional le 4: Figure S1C). Finally, all the unigenes were used for further identi cation of DEGs after the exclusion of abnormal samples.

Identi cation of DEGs associated with alkaline salt stress
To identify the DEGs of the two different genotypes under alkali stress, the expression patterns of DEGs were analysed by comparing 6-h and 24-h libraries with the control library for Alamo and AM-314/MS-155. A total of 10,219 DEGs whose expression was up-or downregulated between samples (fold change ≥ 2 and false discovery rate (FDR) < 0.01) at any pair of alkaline salt-treated points were identi ed (Additional le 6: To further analyse the effects of alkaline salt treatment in the two genotypes over time, the expression trends of DEGs were clustered into 16 modules ( Fig. 2A). Then, the mainstream gene expression trends (6 groups) were screened over the duration of the stress (Fig. 2B). The rst group contained 1,831 enriched genes, which mainly re ected the functional classi cation of genes expressed abundantly in AM-314/MS-number of cytochrome and energy-related genes were enriched in the ribosome and oxidative phosphorylation pathways and gradually downregulated with increasing stress time (Additional le 7: Table S6). The expression level of genes in group 2 (1,492 genes) decreased sharply in the early stage of alkaline salt stress in both switchgrass genotypes. KEGG analysis of the genes in group 2 revealed that most were involved in the ribosome, phenylalanine metabolism, plant hormone signal transduction, and ribosome biogenesis in eukaryotes pathways (Additional le 8: Table S7). The expression of genes in group 3 (1,261 genes) was completely opposite to that in group 1, mainly re ecting the functional classi cation of genes expressed abundantly in Alamo but with little or no expression in AM-314/MS-155. Genes in this group functioned mostly in the glycolysis/gluconeogenesis, plant-pathogen interaction, amino sugar and nucleotide sugar metabolism, and ubiquitin-mediated proteolysis categories (Additional le 9: Table S8). The expression of genes in group 4 (1,072 genes) increased strongly in the early stage of alkaline salt stress, and most of these genes functioned in starch and sucrose metabolism, valine, leucine and isoleucine degradation, plant hormone signal transduction, galactose metabolism, and fatty acid degradation (Additional le 10: Table S9). Although there was a fast decrease in gene expression levels in both group 5 (718 genes) and group 6 (736 genes), the decrease in the expression of genes in group 5 was mainly in AM-314/MS-155, while the decrease in the expression of genes in group 5 was mainly in Alamo. Genes involved in oxidative phosphorylation, protein processing in the endoplasmic reticulum, glycolysis/gluconeogenesis, and glycerophospholipid metabolism were enriched in group 5 (Additional le 11: Table S10), and genes in group 6 were enriched in ribosome, oxidative phosphorylation, glycolysis/gluconeogenesis, and DNA replication (Additional le 12: Table S11).

Identi cation of TFs and PKs
In this study, 1,480 TF genes were predicted using ITAK software (http://itak.feilab.net/cgibin/itak/index.cgi) and were classi ed into 64 TF families (Additional le 13: Table S12). The most abundant TF families were the C2H2 (153 genes), bZIP (108 genes), bHLH (93 genes), C3H (91 genes), MYB-related (82 genes), NAC (81 genes), WRKY (74 genes), GRAS (57 genes) and AP2/ERF-ERF (52 genes) families. The genes in the bHLH, C2H2, NAC and WRKY families were mainly upregulated under alkali stress, while the genes in the AP2/ERF-ERF, bZIP and MYB-related families showed a relatively balanced number of upregulated and downregulated members (Additional le 14: Figure S2). There were 32 TF genes that showed differential expression in both Alamo and AM-314/MS-155 under alkaline salt stress, including 7 TF genes that showed differential expression only at 6 h, 10 TF genes that showed differential expression only at 24 h, and 15 TF genes that showed differential expression at both 6 and 24 h (Additional le 15: Table S13). Interestingly, the TF genes only differentially expressed at 6 h were all downregulated, and most of them belonged to the MYB-related family. Nine of ten TF genes differentially expressed only at 24 h were unregulated, and most of them belonged to the NAC family. There were 27 upregulated TF genes and 16 downregulated TF genes speci cally expressed in Alamo, and many of them were differentially expressed in the late stage of alkali stress (24 h) and belonged to the bZIP, MYBrelated and bHLH families (Additional le 16: Table S14).
In addition, a total of 1,718 PKs, such as calcium/calmodulin-dependent protein kinase (CaMK), CDPK, MAPK and receptor-like kinase (RLK), were predicted in AM-314/MS-155 and Alamo under alkaline salt stress for 6 and 24 h (Additional le 17: Table S15). The CBL-interacting protein kinases (CIPKs), which play a role as Ca 2+ sensors, were upregulated and highly expressed in Alamo under alkaline salt stress for 24 h. RLKs accounted for the largest proportion among many protein kinase families, and the genes belonging to the RLK family maintained a high expression level at 6 h and were then downregulated at 24 h.
Weighted gene coexpression network analysis (WGCNA) of DEGs in response to alkaline salt stress To further study the patterns of association between the DEGs responding to alkaline salt stress and physiological indicators across the two genotypes, WGCNA was performed to explore the gene modules for synergistic expression. After the ltering of low-expression genes (FPKM <5), the remaining 7,485 genes were classi ed into 20 different modules (combined modules) (Fig. 3A).  Figure S3D).
Veri cation of RNA-Seq sequencing data by qRT-PCR analysis The DEGs associated with alkaline salt stress were selected for qRT-PCR assays to verify the accuracy of the RNA-Seq sequencing data. Ten genes were selected randomly from the DEGs that were coexpressed in all sequencing samples. The expression patterns of these ten genes were the same as those of the RNA-Seq assay (Fig. 4A). The expression levels indicated by the qRT-PCR results were strongly correlated with the differential gene expression levels identi ed by mRNA-seq according to the Pearson correlation coe cients, which were 0.8934 and 0.9144 under alkaline salt stress for 6 and 24 h, respectively (Fig.  4B), and demonstrated the reliability of the RNA-seq data.

Discussion
Comparison of physiological differences between different switchgrass genotypes Due to the combined effects of osmotic, ionic and secondary stresses, plant growth is severely affected by alkaline salt stress [7]. Alkaline salt stress causes a large amount of soluble salts, such as Na + , to ow into plant cells, reduces the water potential on the root surface, and reduces the water absorption of plants [8]. Cell membrane permeability has been used as an effective selection criterion for alkaline salt resistance in plants [30]. With the prolongation of alkali stress, the cell membrane is destroyed, resulting in increased membrane permeability and severe lipid peroxidation, and alkali-tolerant genotypes show an obvious advantage [31]. In our study, the ASTTI value of the root RWC showed an upward trend, while the ASTTI values of the REC and MDA content showed a downward trend. Additionally, the root REC and MDA content in AM-314/MS-155 were signi cantly higher than those in Alamo at 4 h post alkaline salt stress. These results indicated that Alamo was more alkaline salt tolerant than AM-314/MS-155, and the degree of damage in Alamo was gradually reduced compared with that of AM-314/MS-155.
The accumulation of osmotic adjustment substances, such as proline, soluble protein and soluble sugar, is an important factor in sustaining plant growth under alkaline salt conditions [32]. The relative alkalitolerant varieties resist alkaline salt stress by enhancing compatible solutes, including proline, thus maintaining cell turgor and water potential to achieve better growth and development [33]. In our study, the accumulation of soluble proteins was relatively lagging, which indicated that proline and soluble sugar could be used as important osmotic regulators in switchgrass under alkali treatment.
Antioxidant enzymes are the most important components of the ROS scavenging system. SOD can catalyse the formation of hydrogen peroxide and oxygen by superoxide radicals, and POD and CAT further scavenge H 2 O 2 [34]. In this paper, the SOD activity of the alkali-tolerant Alamo increased rapidly at 3 h. The expression patterns of POD and SOD induced by alkali stress were similar in switchgrass. The nonenzymatic substance GSH in the antioxidant system also has the effect of scavenging ROS, and the glutathione transferase (GST) family is also a very important component of the metabolic detoxi cation enzyme system [35]. The GSH content of the sensitive genotype AM-314/MS-155 was signi cantly higher than that of the alkali-tolerant genotype Alamo in the early stage of alkaline salt stress. We found that the glutathione S-transferase genes in sensitive cultivars were signi cantly upregulated in the early stage of alkaline salt stress, and this may have been controlled by the regulation of the related genes, leading to changes in physiological indicators. This might be due to the consumption of GST activity, which decreased the GSH content of Alamo. Alternatively, the lower levels of other antioxidants in AM-314/MS-155 were due to the overproduction of GST in competition with other antioxidants. These results revealed the physiological responses of switchgrass to complex alkaline salt stress, including the initial injury accumulation phase (0-6 h) and subsequent gradual recovery of in ltration and ion homeostasis phases (6-24 h).
Gene expression in response to alkaline salt stress in two switchgrass genotypes In this study, a total of 10,219 DEGs were identi ed in two switchgrass genotypes under alkaline salt stress. The proportion of upregulated genes in Alamo was higher than that in AM-314/MS-155, and the DEGs in AM-314/MS-155 were mainly concentrated in the early stage, suggesting quick adjustments in dealing with the damage caused by alkali-salt stress in the short term. All DEGs clustered into 16 modules, and the mainstream gene trends were grouped into six groups (Fig. 2). A large number of ribosomal proteins and a small number of cytochrome and energy-related genes were enriched in the ribosome and oxidative phosphorylation pathways in group 1. These downregulated genes regulate plant growth, development and energy metabolism. Thirteen genes were enriched in the glycolysis/gluconeogenesis pathway, and ve of them were hardly expressed in Alamo. However, there were 9 genes enriched in the glycolysis/gluconeogenesis pathway in group 6. All these genes were upregulated in the alkali-tolerant genotype, and no expression was observed at 6 h and 24 h in the sensitive genotype. Glycolysis/gluconeogenesis was considered to be the key pathway in the process of root development because the component exchanges with other pathways, such as arachidonic acid metabolism, glycan structure degradation, limonene and pinene degradation, and N-glycan degradation, were strongly dependent on its existence [36]. These results suggested that the sensitive genotype was not resistant to alkali stress in the late stage. Some disease resistance proteins were enriched in the plantpathogen interaction pathway in group 6; they all contained the NB-ARC domain, which plays an important role in plant hypersensitivity, and hypersensitivity is a prerequisite for the initiation of defence responses [37,38]. Three ubiquitin-conjugating enzymes and ubiquitin-protein ligase enzymes (Unigene_009730, Unigene_005130, Unigene_018815) were enriched in the ubiquitin-mediated proteolysis pathway, thus degrading proteins damaged by stress through the ubiquitin-protease pathway and maintaining plant homeostasis [39,40].

Identi cation of alkali-tolerant TFs in two switchgrass genotypes
In this study, bHLH, bZIP, C2H2 and WRKY were signi cantly enriched at 6 h in AM-314/MS-155, while bZIP, MYB-related and NAC were signi cantly enriched at 24 h in Alamo. The speci c upregulated TFs associated with high alkali tolerance in Alamo were AP2-ERF, bZIP, bHLH and MYB-related. AP2-ERF transcription factors are involved in many physiological processes, including plant growth [41], abiotic stress responses [42], and ethylene and abscisic acid signalling pathways [43]. It was reported that Gm-ERF3 isolated from soybean could improve the salt tolerance of transgenic tomato [44]. Overexpression of PvERF001 isolated from the ERF genes of switchgrass could also increase biomass yield and sugar release e ciency in transgenic lines [45]. In this paper, we observed that AP2-ERF family TFs were still signi cantly upregulated at 24 h and might be used as candidate TFs to improve alkali tolerance in switchgrass. Overexpression of bZIP TFs can signi cantly increase plant salt tolerance [46,47]. In our study, the bZIP TFs continuously responded positively to alkaline salt stress at 6 h and 24 h and, combined with the ABA responsive element binding protein (AREB) or ABRE binding factors (ABF) [48], activated the ABA-dependent signal transduction pathway, participated in regulating the expression of ABA-related genes [49], and enhanced plant resistance to environmental stress. We also found that these TFs were involved in blue light regulation in the circadian rhythm-plant pathway, which affects the photomorphological regulation of plants [50]. Overexpression of bHLH can help plants positively respond to salt stress independent of the ABA signalling pathway [51]. In our study, bHLH TFs were speci cally upregulated at 6 and 24 h in Alamo, indicating the alkaline salt tolerance of Alamo. Many MYB TFs are involved in plant responses to adverse environments. Rong et al showed that OsMYBc in rice can regulate OsHKT1 to deal with Na + damage caused by salt stress and prevent leaf accumulation of Na + toxicity [52]. An MYB-related gene, LcMYB1, identi ed from Leymus chinensis was found to be involved in the response to high salinity, drought and abscisic acid in transgenic Arabidopsis and was expressed at higher levels in the rhizome and panicle [53]. In our study, multiple MYB TFs were highly expressed at 6 and 24 h and could be considered target genes involved in the alkali stress response.

Identi cation of functional genes and potential regulatory response mechanisms of alkali tolerance in switchgrass under alkaline salt stress
Under alkaline salt stress, the genes encoded transporters responsible for reconstituting osmotic and ionbalanced proteins under alkali stress, such as sodium/hydrogen exchange protein (NHX), alkaline and neutral invertase (CINV2), and heat shock proteins (HSPs) including HSP20, HSP70, Hsp90 and DanJ, were differentially expressed in the two switchgrass genotypes. The ABC transporter and H + transport ATPase genes were signi cantly upregulated at 6 and 24 h in the two genotypes and showed a downward trend in Alamo in the late stage. At the same time, osmotic and ionic stresses can also lead to an increase in the concentration of free Ca 2+ in the cytosol, which results in downstream gene phosphorylation and protein cascade [54]. The CDPK and MAPK pathways are important signalling pathways in plant responses to salt stress [55]. The MAPK cascade is an important regulator of antioxidant defence [56]. CAMS/CMLS, CIPKs and CDPKs, which function as Ca 2+ sensors, are involved in many growth, development and stress-induced signal transduction pathways [57,58]. Overexpression of OsCIPK1 and OsCIPK9 not only occurs in response to salt stress but also to ABA, drought and cold stress [59]. In this study, MAPK6 was signi cantly upregulated at 6 and 24 h in the two genotypes. CIPKs were mainly upregulated at 6 and 24 h in Alamo, while expression in AM-314/MS-155 remained low. These results suggest that the CDPK and MAPK signalling pathways may also mediate the response of switchgrass to alkali stress.
Compared with the two genotypes, many more DEGs were involved in plant hormone signal transduction in AM-314/MS-155 than in Alamo at 6 h, while more DEGs were identi ed in Alamo at 24 h. Alamo showed differential expression in the ABA signalling pathway at two different time points, while AM-314/MS-155 showed no signi cantly enriched DEGs. ABA plays a very important role in plant salttolerance signal transduction pathways and participates in the physiological and biochemical processes of abiotic stress in plants [60]. In our study, the HVA22 protein, a unique ABA stress-induced protein, was mainly upregulated in Alamo at 24 h and in AM-414/MS-155 at 6 h under alkaline salt stress. The HVA22 protein is induced by environmental stresses such as drought and high salt, which could inhibit GAmediated programmed cell death of cereal aleurone cells and improve plant resistance to stress [61].
ROS also act as second messengers [62,63]. Detoxi cation proteins involved in ROS scavenging were highly expressed at 6 and 24 h. Some DEGs were expressed more in AM/314/MS-155 than in Alamo, but most of the favourable upregulated genes were speci cally expressed in the alkali-tolerant genotype Alamo. LEA proteins have high hydrophilicity and can supplement cells with su cient water during stress and lessen the damage caused by osmotic stress [64]. Dehydrin, a second type of LEA protein, can signi cantly increase the resistance of plants to high salt and osmotic stress [65]. In this paper, the LEA protein was downregulated in AM-314/MS-155 but upregulated in Alamo. The dehydrin protein showed speci c upregulation only in Alamo, which fully re ected the positive response of Alamo to alkali stress.
The alkaline/neutral sucrase genes were also upregulated at 6 and 24 h under alkaline salt stress, and the number and expression of genes upregulated in Alamo were higher than those in AM-314/MS-155. Alkaline/neutral sucrase is a class of genes unique to plants and photosynthetic bacteria. The total alkaline/neutral invertase activity in wheat leaves increased after osmotic stress or low temperature stress, resulting in increased expression of related genes [66].
These genes are involved in antioxidant systems and transport functions and can aid in plant adaptation to alkaline saline environments by regulating ion homeostasis, transport proteins, detoxi cation, heat shock proteins, dehydration and sugar metabolism. The interaction of these genes constituted a complex regulatory network (Fig. 5). We hypothesized that switchgrass plants respond to alkaline salt stress through ion and osmotic stress signals. These signals were rst sensed by receptors present on the plant cell membrane and then involved in signal transduction through ion channels and carrier proteins. The receptors in the roots may recognize salt signals, causing an instantaneous increase in ROS, leading to the opening of some plant hormone signalling pathways (including those of auxins, abscisic acid, cytokinins, gibberellins, ethylene, jasmonic acid and salicylic acid), biosynthesis of ubiquinone and other terpenoid-quinones, phenylpropanoid metabolism and biosynthetic pathways.

Conclusions
Taken together, we investigated the alkaline salt stress tolerance mechanism of switchgrass by analysing

Alkaline salt stress treatments
Alkaline salt stress treatments were conducted when the seedlings reached the E5 developmental stage [68]. The alkaline salt solution, which was made of Na 2 CO 3 and NaHCO 3 (1:9, v/v), with Na + at 150 mM and pH 8.9, was prepared with half-strength Hoagland's nutrient solution. Roots of the two genotypes were harvested at 0, 3, 6, 12, and 24 h post alkaline salt stress. The experiment was repeated three times.
For each biological replicate, some of the samples were used to determine the relative water content (RWC) and relative electric conductivity (REC) directly, and the rest were frozen in liquid nitrogen immediately and stored at -80 °C for physiological and transcriptome analyses.
ASTTI evaluation by physiological assays Physiological traits including the RWC, REC, the contents of MDA, free proline, soluble protein, soluble sugar, and reduced GSH, and the activities of SOD, POD and CAT were assayed three times with three biological replicates [69][70][71][72]. The ASTTI, which was calculated based on the formula: ASTTI = (value of a trait under salt stress condition)/(value of a trait under controlled condition), was used to estimate the alkaline salt tolerance of the two switchgrass genotypes [73,74].
RNA extraction, library preparation and next-generation sequencing (NGS) analysis Total RNA from each sample used for transcriptome analysis was extracted with TRIzol reagent (Invitrogen, USA) according to the manufacturer's protocol, and genomic DNA was removed via digestion with DNase I (TaKaRa, Japan). The RNA sample concentration and quality were determined using a Nanophotometer P330 spectrophotometer (IMPLEN, Germany). A total amount of 3 μg RNA per sample was used as input material for the RNA sample preparations. Sequencing libraries were generated using the NEBNext®Ultra™ RNA Library Prep Kit for Illumina® (NEB, USA) following the manufacturer's recommendations, and index codes were added to attribute sequences to each sample.
The clustering of the index-coded samples was performed on a cBot Cluster Generation System using TruSeq PE Cluster Kit v3-cBot-HS (Illumina) according to the manufacturer's instructions. After cluster generation, the library preparations were sequenced on an Illumina HiSeq 2000 platform, and paired-end reads were generated. Clean reads were obtained by removing reads containing adapters, reads containing poly-N and low-quality reads from the raw data. All downstream analyses were based on clean data with high quality. Transcriptome assembly was accomplished using Trinity Software with min_kmer_cov set to 2 by default and all other parameters set to the default selections [75]. The clean reads of each sample were sequence aligned with the assembled unigene library, and the transcriptome sequencing library was evaluated for quality. Gene function was annotated by using BLAST software based on the NR, Pfam, KOG, COG, eggNOG, Swiss-Prot, KEGG and GO databases [76].

Identi cation of DEGs, putative TFs and PKs
The paired reads were compared with the unigene library using Bowtie [77] and RNA-seq by expectation maximization (RSEM) [78], and the expression abundance of a unigene was expressed using the FPKM value [79]. Differential expression analysis under the two conditions was performed using the DESeq R package (1.10.1). A false discovery rate (FDR) <0.01 and fold change (FC) ≥ 2 were used as screening criteria, and samples with a strong correlation between samples were selected for subsequent functional expression annotation and enrichment analysis of the DEGs. GO enrichment analysis of the DEGs was implemented by the topGO R package-based Kolmogorov-Smirnov test. KOBAS software was subsequently used to test the statistical enrichment of DEGs in the KEGG pathways [80]. TFs and PKs were predicted using ITAK software (http://itak.feilab.net/cgi-bin/itak/index.cgi).
Coexpression network analysis with WGCNA Coexpression networks were constructed via the WGCNA package in R from all the DEGs [81]. Modules were obtained by the automatic network construction function using blockwise modules with the default settings. The eigengene value was calculated for each module and used to test the association with each physiological index. The total connectivity and intramodular connectivity (function soft connectivity), kME (for modular membership), and kME p values were calculated for the DEGs [82].  times with three biological replicates. The data, shown as means ± SEs, were subjected to student's t test to determine signi cant differences. "*" means p<0.05, "**" means p<0.01.   Veri cation of RNA-Seq sequencing data by qRT-PCR assay (A) and Pearson correlation coe cients of genes under alkali-salt stress for 6 and 24 h (B). The y-axis represented qRT-PCR relative expression levels from three biological replicates and the log2 fold-change of the unigenes.