Transcriptome Changes Induced by Different Potassium Levels in Banana Roots

Potassium plays an important role in enhancing plant resistance to biological and abiotic stresses and improving fruit quality. To study the effect of potassium nutrient levels on banana root growth and its regulation mechanism, four potassium concentrations were designed to treat banana roots from no potassium to high potassium. The results indicated that K2 (3 mmol/L K2SO4) treatment was a relatively normal potassium concentration for the growth of banana root, and too high or too low potassium concentration was not conducive to the growth of banana root. By comparing the transcriptome data in each treatment in pairs, 4454 differentially expressed genes were obtained. There were obvious differences in gene function enrichment in root systems treated with different concentrations of potassium. Six significant expression profiles (profile 0, 1, 2, 7, 9 and 13) were identified by STEM analysis. The hub genes were FKF1, HsP70-1, NRT1/PTR5, CRY1, and ZIP11 in the profile 0; CYP51 in profile 1; SOS1 in profile 7; THA, LKR/SDH, MCC, C4H, CHI, F3′H, 2 PR1s, BSP, TLP, ICS, RO, chitinase and peroxidase in profile 9. Our results provide a comprehensive and systematic analysis of the gene regulation network in banana roots under different potassium stress.


Introduction
Potassium is a highly mobile nutrient mass that is an integral part of many physiological and biochemical processes in plants [1]. Potassium is essential for many cell and tissue activities, such as osmoregulation, enzyme activity, transport of minerals and metabolites, and stomatal aperture regulation [2][3][4][5][6][7]. It is involved in programmed cell death and senescence [8,9]. It also plays important roles in increasing plant resistance to disease [10][11][12][13]. Potassium deficiency can reduce plant photosynthesis, transpiration rate and stomatal conductance, and affect crop growth, yield quality and resistance to biological and abiotic stresses [7,14]. When plants face nutrient starvation in the soil, they will adjust the structure of root system and use the remaining resources to find a more favorable soil environment [15]. Arabidopsis thaliana significantly reduced lateral root elongation under low potassium treatment [16,17]. Maize varieties with high potassium uptake rate had advantages over those with low potassium uptake rate in terms of root length, root volume and root surface

Overview of Transcriptome Sequencing Results
To overview the banana roots gene expression patterns at different potassium supply conditions, we carried out the RNA-seq analysis. Seedling root samples treated as K0, K1, K2 and K3, were sequenced. Approximately 24.02 to 34.47 million 150 bp paired-end clean reads were obtained from K0, K1, K2 and K3 samples, respectively. The average Q20 contents were 94.01% to 94.15%, and GC contents were 53.58% to 54.26%, respectively. These data indicate that the sequencing results were fine. The mapped reads information among K0, K1, K2 and K3 samples was: 77.27% to 83.37% of the total mapping ratio ( Table 1). The transcriptome data was qualified and considered appropriate for subsequent analysis.

Different Concentration of Potassium Stresses Affect Genes Expression in Banana Roots
More than 68% of the genes had gene expression levels of FPKM ≥ 2 in each evaluated instance. Among the above gene clusters, more than 70% of genes were FPKM < 25, and more than 25% had FPKM between 10 and 25 (Table S1). About 3% of genes had FPKM values more than 200. It is clear that the gene expression levels in the four treatments displayed the same trend ( Figure 2A). The gene expression distribution analysis showed that among the 24,277 genes with FPKM ≥ 2, 19,872 (82%) genes were expressed in all four potassium treatments, while a small number were expressed only in one condition (2.8% of K0, 1.5% of K1, 0.8% of K2, and 1.3% of K3) ( Figure 2B).

Overview of Transcriptome Sequencing Results
To overview the banana roots gene expression patterns at different potassium supply conditions, we carried out the RNA-seq analysis. Seedling root samples treated as K0, K1, K2 and K3, were sequenced. Approximately 24.02 to 34.47 million 150 bp paired-end clean reads were obtained from K0, K1, K2 and K3 samples, respectively. The average Q20 contents were 94.01% to 94.15%, and GC contents were 53.58% to 54.26%, respectively. These data indicate that the sequencing results were fine. The mapped reads information among K0, K1, K2 and K3 samples was: 77.27% to 83.37% of the total mapping ratio ( Table 1). The transcriptome data was qualified and considered appropriate for subsequent analysis.  The horizontal orange bar graph on the left side represents the element statistics of each set, the single black point in the middle matrix represents the element unique to a set, the line between points and points represents the intersection unique to different sets, and the vertical blue bar graph represents the corresponding intersection element values.

Gene function Enrichment Analysis of DEGs in Responses to K + Treatments
Gene ontology (GO) analysis was conducted to elucidate the molecular function of the DEGs detected in banana roots from K1, K2 and K3 compared to K0 using the OMICSHARE online program (Tables S2). Many GO terms were significantly enriched (adjusted p < 0.05) and the GO terms enriched in three kind potassium treatments were different (Tables S2). 21 GO terms were enriched in K1 upregulated DEGs, while 7 GO terms were significantly enriched in K1 down-regulated DEGs. 9 GO   The horizontal orange bar graph on the left side represents the element statistics of each set, the single black point in the middle matrix represents the element unique to a set, the line between points and points represents the intersection unique to different sets, and the vertical blue bar graph represents the corresponding intersection element values.

Gene function Enrichment Analysis of DEGs in Responses to K + Treatments
Gene ontology (GO) analysis was conducted to elucidate the molecular function of the DEGs detected in banana roots from K1, K2 and K3 compared to K0 using the OMICSHARE online program (Tables S2). Many GO terms were significantly enriched (adjusted p < 0.05) and the GO terms enriched in three kind potassium treatments were different (Tables S2). 21 GO terms were enriched in K1 upregulated DEGs, while 7 GO terms were significantly enriched in K1 down-regulated DEGs. 9 GO

Gene function Enrichment Analysis of DEGs in Responses to K + Treatments
Gene ontology (GO) analysis was conducted to elucidate the molecular function of the DEGs detected in banana roots from K1, K2 and K3 compared to K0 using the OMICSHARE online program (Table S2). Many GO terms were significantly enriched (adjusted p < 0.05) and the GO terms enriched in three kind potassium treatments were different (Table S2). 21 GO terms were enriched in K1 up-regulated DEGs, while 7 GO terms were significantly enriched in K1 down-regulated DEGs. 9 GO terms in K2 up-regulated DEGs, while 13 GO terms were enriched in K2 down-regulated DEGs. 5 GO terms were enriched in K3 up-regulated DEGs, while 19 GO terms were enriched in K3 down-regulated DEGs. In K1 up-regulated DEGs, the "glycopeptide alpha-N-acetylgalactosaminidase activity", "manganese ion binding", "nutrient reservoir activity", "transporter activity" related, "ATPase activity" related, "hydrolase activity, acting on acid anhydrides, catalyzing transmembrane movement of substances" related, and "transferase activity, transferring acyl groups" GO terms were specific enriched. In K1 down-regulated DEGs, the "cysteamine dioxygenase activity", "xyloglucan:xyloglucosyl transferase activity", "dioxygenase activity", and "hydrolase activity, acting on glycosyl bonds" GO terms were specific enriched In K2 up-regulated DEGs, the "linoleate 13S-lipoxygenase activity" GO term was specific enriched. In K2 down regulated DEGs, the "sequence-specific DNA binding" GO term was specific enriched. In K3 down-regulated DEGs, the "methylcrotonoyl-CoA carboxylase activity", "chitin binding", "peroxidase activity", "oxidoreductase activity, acting on peroxide as acceptor", and "antioxidant activity" GO terms were specific enriched. While in K3 up-regulated DEGs, no GO term was specific enriched (Figure 4, Table S2). These result suggested that banana roots require genes with different functions to survive with different potassium concentrations.

Expression Trend Analysis
The expression profiles of 4454 DEGs were determined and 20 candidate profiles were obtained ( Figure 5A), in which six significant (p < 0.05) expression profiles (profile 0, 1, 2, 7, 9 and 13) were identified. The 537 DEGs in Profile 0 were significantly present down-regulated with the potassium concentration increase. Profile 1 included 393 DEGs that were down-regulated from K0 to K2 and up-regulated from K2 to K3, which is consistent with the phenotype of the banana root growth. Profile 2 contained 241 DEGs down-regulated from K0 to K1 and present no change from K1 to K3. The 382 DEGs in Profile 7 presented no change from K0 and K1, K2 and K3, but down regulated from K1 to K2. In profile 9, the 596 DEGs present no change in K0 to K2, but down-regulated from K2 to K3. In profile 13, 314 DEGs up-regulated from K0 to K1, then down-regulated to K3 ( Figure 5). regulated from K2 to K3, which is consistent with the phenotype of the banana root growth. Profile 2 contained 241 DEGs down-regulated from K0 to K1 and present no change from K1 to K3. The 382 DEGs in Profile 7 presented no change from K0 and K1, K2 and K3, but down regulated from K1 to K2. In profile 9, the 596 DEGs present no change in K0 to K2, but down-regulated from K2 to K3. In profile 13, 314 DEGs up-regulated from K0 to K1, then down-regulated to K3 ( Figure 5).

GO Enrichment Analysis of the Six Significant Gene Expression Patterns
To understand the functional enrichment of DEGs in the six significant profiles, we conducted GO analysis on them respectively, and regarded the adjusted p value < 0.05 as significant enrichment. There were 4, 1, 19, 21, and 2 GO terms significantly enriched in profiles 0, 1, 7, 9 and 13, and zero enriched in profile 2. Significant GO terms that were related to ion binding such as iron ion binding, transition metal ion binding, and manganese ion binding were enriched in profiles 7, 9, and 13. Significant GO terms such as transcription factor activity were enriched in profiles 0 and 7; oxidoreductase activity related GO terms were enriched in profiles 1, 7, and 9 ( Table 2). These results suggest that the banana root has evolved a range of molecular strategies that response in different

GO Enrichment Analysis of the Six Significant Gene Expression Patterns
To understand the functional enrichment of DEGs in the six significant profiles, we conducted GO analysis on them respectively, and regarded the adjusted p value < 0.05 as significant enrichment. There were 4, 1, 19, 21, and 2 GO terms significantly enriched in profiles 0, 1, 7, 9 and 13, and zero enriched in profile 2. Significant GO terms that were related to ion binding such as iron ion binding, transition metal ion binding, and manganese ion binding were enriched in profiles 7, 9, and 13. Significant GO terms such as transcription factor activity were enriched in profiles 0 and 7; oxidoreductase activity related GO terms were enriched in profiles 1, 7, and 9 ( Table 2). These results suggest that the banana root has evolved a range of molecular strategies that response in different potassium concentrations.

Validation of RNA Sequencing Data
Quantitative real-time PCR (qRT-PCR) were performed to evaluate the accuracy of the gene expression in this transcriptome data. In this study, eight DEGs were randomly selected to detected by qRT-PCR. The expression patterns analysis showed that the qRT-PCR results were consistent with the transcriptome patterns. This result indicated that the RNA-Seq data was reliable (Figure 7).

Validation of RNA Sequencing Data
Quantitative real-time PCR (qRT-PCR) were performed to evaluate the accuracy of the gene expression in this transcriptome data. In this study, eight DEGs were randomly selected to detected by qRT-PCR. The expression patterns analysis showed that the qRT-PCR results were consistent with the transcriptome patterns. This result indicated that the RNA-Seq data was reliable (Figure 7).

Discussion
Plant roots can sense changes in the external environment and adjust the absorptive capacity by changing root morphology [30]. The length and surface area of a root system can be enhanced in potassium-rich soil, while the length and surface area of root system can be restricted in potassiumdeficient soil [31][32][33]. Thus, it was not surprising to detect the inhibition of banana root hair growth in potassium deficiency (K0) and low potassium (K1) treatments, in the present study. As it was rarely

Discussion
Plant roots can sense changes in the external environment and adjust the absorptive capacity by changing root morphology [30]. The length and surface area of a root system can be enhanced in potassium-rich soil, while the length and surface area of root system can be restricted in potassium-deficient soil [31][32][33]. Thus, it was not surprising to detect the inhibition of banana root hair growth in potassium deficiency (K0) and low potassium (K1) treatments, in the present study. As it was rarely reported the regulation of high potassium levels on plant root growth, it was meaningful to find out that the banana root length increased and root hairs were inhibited under the high potassium (K3) treatment. With the purpose to finding out the mechanism of different potassium levels in regulating the banana root growth, we performed the RNA-seq transcriptome analysis. The expression profile analysis of DEGs showed six significantly profiles, including profile 0, 1, 2, 7, 9 and 13 ( Figure 6). Coincidentally, the expression trend of 393 DEGs in profile 1 was contrary with the root growth phenotype change trend treated by different levels of potassium. These genes were up-regulated by potassium stresses compared with normal potassium level, and it is speculated that they were the main genes leading the change of root phenotype. GO enrichment analysis showed that one GO term (including 17 DEGs), which is related to oxidoreductase activity, was significantly enriched in profile 1 ( Table 2). The 17 DEGs included nine Cytochrome P450 family members, four oxidoreductase family members, two 1-aminocyclopropane-1-carboxylate oxidase family members, one Omega-6 fatty acid desaturase, and one S-norcoclaurine synthase (Table S15). The oxidoreductase activity enrichment also had reported in other plants treated by low nutrient stresses, such as in low potassium treated pear fruit [34], in low nitrogen treated wheat [35], in low phosphorus treated sorghum [36]. The networks of the DEGs in profile 1 were conducted, and the results revealed that CYP51, a member of cytochrome P450 family was the hub gene with the highest node degree 8 ( Figure 6B, Tables S5 and S6). In low potassium treated pear fruit, there multiple cytochrome P450 members were detected [34]. In low potassium treated tobacco, 3 cytochrome P450 members were also up-regulated [37]. The DEGs belong to cytochrome P450 family were also detected from potassium deficiency tomato [38], low-N polar [39], rice [40], sorghum [41] and wild barley [42]. Cytochrome P450 family genes were reported to participate in the biosynthesis of lignin precursors or hormone homeostasis [43], involved in the formation of cell wall and regulated the plant growth [44]. Therefore, the focus of CYP51 gene in DEGs interaction network of profile 1, indicated a significant correlation with root morphological changes under potassium stresses.
In profile 7, the DEGs were high regulated at low potassium levels (K0, K1) compared to K2 and K3 ( Figure 6). GO analysis showed that the ion binding, transporter activity, Kinase, oxidoreductase activity, Transcription factors related DEGs were significantly enriched in profile 7 ( Table 2). The DEGs interaction network indicated that SOS1 was the hub gene with 10 nodes (Figure 6C, Tables S7 and S8). SOS1 is a plasma membrane Na + /H + antiporter and belongs to the "salt overly sensitive" (SOS) signaling pathway [45]. It was reported to be an important gene locus for salt tolerance and potassium acquisition in plants [46][47][48]. Previous studies showed that SOS1 plays an essential function in regulating the spatial distribution and long-distance transportation of Na + and K + , the transport system of K + and Na + , keeping the Na + and K + homeostasis, and maintaining the normal growth of plants [49]. Here, the SOS1 was at a pivotal position in the interaction network of profile 7 DEGs, indicating that it played a key role in regulating banana root response to potassium stress. Early studies revealed that the CBL1/CBL9-CIPK23-AKT1 module is the primary K + uptake regulatory pathway to plants response in potassium starvation [50]. However, in our study, the AKT1, CBL1, and CBL9 were not present in the interacting networks of profile 7. Then we dug into the transcriptome data and found that they were also not present in 4454 DEGs. The results suggested that the CBL1/CBL9-CIPK23-AKT1 pathway was not involved in banana root system in response to potassium stress. AtKC1 is a K + channel regulatory subunit, which together with AtCIPK23, has been reported to coordinate the regulation of AtAKT1-mediated K + short stress response in Arabidopsis [51,52]. In addition, the CBL4/CIPK6 complex was revealed to modulating AKT2 activity in Arabidopsis [53]. CBL4 (SOS3) interacted with SOS2, and further activated SOS1 by phosphorylation [54][55][56]. TPK1, a voltage independent, Ca 2+ /pH-activated two-pore K + channel, was proved playing key functions in keeping intracellular K + homeostasis [57,58]. KUP3 belongs to the largest potassium transporter HAK/KUP/KT family, which was strong inducted by potassium starvation [59]. In this study, it was predicted that KC1 interacted with CIPK23, TPK1 and KUP3 respectively, and CIPK23 also interacted with SOS1 and CBL4(SOS3) respectively ( Figure 6C). Given that the interaction between SOS1 and SOS3 [56], KC1 and CIPK23 [51,52], CBL4(SOS3) and CIPK6 [53] has been proved, thus the consistent expression and their interaction of SOS1, KC1, CIPK23, CBL4, CIPK6, KUP3, TPK1 in the network, provide new possible mechanisms for banana roots to response to potassium stress. In a word, the genes in profile 7 were mainly related to ion transport of potassium.
The 537 DEGs in Profile 0 were significantly down-regulated with the potassium concentration increase. In the profile 0 PPI network, five genes, including FKF1, HsP70-1, NRT1/PTR5, CRY1, and ZIP11, were strongly connected to other genes, indicated that they were hub genes in profile 0 ( Figure 6A). The five hub genes were connected by several transcription factors including MYBs and bHLHs. There were also many HsP transcription factors in the network ( Figure 6A, Tables S3 and S4). A recent study showed that cryptochrome 1 (CRY1, plant blue light photoreceptor) actively participates in the formation of secondary cell walls. When CRY1 gene was missing in plants, it will lead to the decrease of cellulose and lignin accumulation in plant stems [60]. Later, FLAVIN-BINDING KELCH REPEAT, F-BOX 1 (FKF1), another blue light receptor, was identified from cellulose synthase-deficient dwarf mutant, and revealed its role in negatively regulation of cellulose biosynthesis [61]. In this study, the hub location of CRY1 and FKF1 in the interaction network of profile 0, indicated that the banana roots might change their cell wall component to response to different potassium stress. In this network, ZIP11, a zinc transporter gene, existed in the hub location. Members of the ZIP family were considered to be important transporters of Zn, Fe, Mn, Cu and other bivalent metal ions [62][63][64]. Previous studies indicated that the relationship between Zinc and potassium in plants is mutual promotion. As the content of Zn in the body increases, the content of K also increases in cotton [65]. ZIP11 has the function of separating Zn/Cd into vesicles and holding back its transfer from root to bud [66]. In addition, there were also some other ZIP or metal transport related genes such as IRT1, HMA2, ZIP5 in this interaction network. Therefore, the different potassium stress in banana roots might cause different zinc or other metal levels in roots through the ZIPs regulating network. A NRT1/PTR family gene, known for specific nitrate and auxin uptake, was found in the hub location of the interaction network of profile 0. It is well known that the nitrogen and potassium fertilizers are interdependent. The expression of several NRT1/PTR genes regulated by external potassium has also been reported on tomato [67], and Arabidopsis [16]. In addition, it has been mentioned that NRT1.1 plays a key role in the K + translocation of Arabidopsis thaliana [68]. The present results were consistent with previous studies, indicating that the exogenous potassium stress can also affect the absorption and transport of nitrogen in banana root. Plant heat shock protein (Hsps) is a kind of protein that can be activated and produced in large quantities when plants encounter abiotic stress. Hsps has the function of molecular chaperone. It can maintain the stability of protein and cell membrane by assisting protein folding, assembly, translocation and degradation, and protect the normal growth of plants [69]. The HSP70 expression has been paid attention in plant response to abiotic stress, for instance drought [70,71] and salt [35,72]. However, its expression in plants response to potassium stress, was rarely mentioned. Nevertheless, its role in human potassium channels has been proven [72,73]. According to the above analysis, genes in profile 0 were mainly related to regulating cell wall formation, other nutrition transport and the Hsps expression.
In profile 9, the 596 DEGs present no change from K0 to K2, but down-regulated from K2 to K3. In profile 9, THA, LKR/SDH, MCC, C4H, CHI, F3 H, PR1s, BSP, TLP, ICS, RO, chitinase, peroxidase were the hub genes with the node degree more than 8 ( Figure 6D, Tables S9 and S10). THA, LKR/SDH, MCC, C4H, CHI, F3 H, ICS, chitinase and peroxidase belong to the metabolic pathways. Threonine aldolase (THA) is involved in the metabolic process of threonine. Threonine is lysed into glycine and acetic acid under the action of threonine aldolase. The results showed that glycine, under the action of glycine methyltransferase, synthesized betaine through three steps of methylation, thus improving the stress resistance of plants [74]. THA may be an important step in the regulation of amino acid metabolism in banana roots under adverse potassium conditions. LKR/SDH (Lysine ketoglutaric acid reductase/saccharopine dehydropine dehydrogenase) participating in the Lys degradation, is found to be a bifunctional enzyme [75]. The expression of LKR/SDH was affected by lysine, osmotic stress, abscisic acid, sugar, jasmonate, nitrogen, salt and drought stresses [76][77][78]. This study indicated that, the expression of LKR/SDH was depressed by high concentration of potassium. Leucine is a branched amino acid involved in the breakdown of large amounts of protein, which accumulates instantaneously in carbon-hungry cells. Methylcrotonoyl-CoA carboxylase (MCC), a mitochondria-localized carboxylase, is involved in the leucine catabolizing pathway [79][80][81]. It is expressed in a large number of sycamore cells during starvation [81,82]. Therefore, MCCase has been proposed as a new biochemical marker of carbohydrate starvation autophagy process [79]. Knockout of MCCa or MCCb results in impaired reproductive growth phenotypes in Arabidopsis thaliana [82]. It was also reported that MCC plays an important role in regulating triglyceride accumulation in Phaeodactylum tricornutum [83]. During cotyledon development, MCCase mRNA accumulation was positively correlated with the increase of organ age [80]. Biotinylation, a post-translational modification, is important for the activity of MCC, which has been demonstrated in tomato [84]. The expression profile of MCC in banana roots treated by different potassium levels, which provided new evidence for its molecular regulatory function. Cinnamate 4-hydroxylase (C4H, EC 1.14.13.11) is one of the key enzymes in the phenylpropanoid pathway, which is also the first Cyt P450-dependent monooxygenase of the pathway [85]. The expression of C4H can be regulated by various environmental stresses. In Arabidopsis [86], Dryopteris fragrans [87], and Salvia miltiorrhiza [88], C4H was induced by light, UVB. In rice [89], orange [90], Carthamus tinctorius [91] and kenafcold [92], C4H was induced by NaCl, cold, H 2 O 2 , ABA and salicylic acid (SA). In Capsicum [93], C4H was induced by drought stress. Chalcone isomerase (CHI; EC 5.5.1.6) catalyze conversion of chalcone to flavonoids to produce most major flavonoid subgroups [94]. The correlation between CHI activity and plant response to environmental stress has been investigated. For example, the addition of fungal inducers led to rapid accumulation of CHI mRNA in bean cell culture [95]; overexpressing GmCHI1A can decrease the harm by Phytophthora sojae in soybean [96]. Similarly, the transcription of CHI gene was induced by salt stress [97][98][99]. F3 H belongs to the cytochrome P450 subfamily and catalyzes the NADPH-dependent monooxygenase reactions [100]. It also participates in the flavonoid biosynthetic pathway. Our work showed that C4H, CHI, and F3 H in banana root were significantly down-regulated under high level potassium stress. In addition, they were also in the hub of the interaction network. It indicates that high potassium stress may inhibit the flavonoid biosynthesis pathway dominated by C4H, CHI and F3 H genes in banana roots. Peroxidase (POD, EC 1.11.1.7) is a recognized enzyme that affects plant growth through regulation of the lignin accumulation [101,102]. The increase of ion-binding POD activity inhibited root growth of rice [103]. By adding substances that inhibit pod activity, cell wall rigidity caused by pod can be removed and root elongation can be promoted [104]. Therefore, the down-regulation of peroxidase gene expression in profile 9 under high potassium treatment, might be the main reason which cause the banana root length increase. It is well known that PR1, chitinases, thaumatin-like proteins, peroxidases have been classified as PR proteins, which are associated with plant defense [105]. Isochorismate synthase 1 (ICS1) is an essential enzyme in the synthesis of salicylic acid (SA). The SA can be synthesized by isochorismate synthase (ICS) pathway, and also can be synthesized by phenylalanine ammonia-lyase (PAL) pathway in plants [106,107]. In Arabidopsis, PAL pathway is mainly for the basic production of SA, while ICS pathway is for the production of pathogen-induced SA [108][109][110]. The accumulation of SA can induce the expression of the pathogenesis-related gene (PR1), which is considered to be an SA receptor in plants [111][112][113][114]. Our study showed that the chitinases, thaumatin-like proteins, peroxidases, ICS1, and 2 PR1 genes were down regulated under high level potassium stress. As they were located in the core location, therefore, it was speculated that high concentration of potassium inhibited their expression and would reduce the disease resistance of banana roots. In conclusion, the genes in profile 9 were suppressed by high potassium, and these genes mainly involved in the metabolic function and plant defense.

Plant Materials and Different Potassium Treatments
One-month-old young Brazilian bananas were washed and transplanted into plastic containers filled with vermiculite and grown in a greenhouse. The culture condition was 28 ± 1 • C, 16 h/d light, 60-70% relative humidity. The plants were wet thoroughly by special nutrient solution every 5 days. Basic nutrient solution was made up of 4 mmol/L Ca(NO 3 ) 2 ·4H 2 O, 3 mmol/L NH 4 NO 3 , 1 mmol/L NH 4 H 2 PO 4 , 2 mmol/L MgSO 4 ·7H 2 O, iron salts and micronutrient were formulated with Hoagland nutrient solution. Four potassium levels from no potassium to high potassium were set to add 0 (K0), 0.03 (K1), 3 (K2), 200 (K3) mmol/L K 2 SO 4 respectively. After four weeks of potassium stress, banana roots were observed, photographed and sampled. The potassium stress treatment of banana root with different concentration was repeated three times.

Extraction of Total RNA from Banana Root under Different Potassium Stresses
Total RNA of banana root treated with different potassium was extracted according to the instructions of Trizol Kit (Promega, USA), after that the RNase-free DNase I (Takara Bio, Shiga, Japan) was added to eliminate residual DNA. The 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA, USA) was used to checked the RNA quality and concentration. Equal amounts of RNA from each sampled tissue were mixed for the subsequent steps of our experiments. Three biological replicates were performed and each biological replicate was constituted by a pool of three plants.

RNA-Seq Library Preparation, Sequencing and Data Analysis
The construction method of the library was the same as that reported by Balic et al. [115]. After being purified and enriched, the 200 bp ± 25 bp cDNA fragments were used to construct cDNA library. The cDNA library sequencing was performed by Gene Denovo Biotechnology Co. (Guangzhou, China) using the Illumina HiSeq™ 2000. Then the obtained sequences were checked to delete the low-quality sequences. In addition, the retained high-quality sequences were assembled. Sequencing reads were mapped to banana genome sequences (https://banana-genome-hub.southgreen.fr/) by SOAPaligner/soap2. The uniquely mapped reads were used to evaluate the expression level. The formula of RPKM= 106C/NL/103 was used to evaluate the expression level as described before [116]. The RPKM value is then used for pairwise comparisons between samples. The gene expression meeting the condition of p-value < 0.05, log2FC ≥ 1 were determined to be significantly different expression. The raw data was uploaded to NCBI, and the BioProject ID is PRJNA589855.

Data Statistics, GO Enrichment, Expression Trend Analysis and Interactive Network Construction of DEGs
The bar chart is drawn on an EXCEL spreadsheet. Analysis of Venn, upset Venn, GO enrichment and gene expression trend were drawn by appropriate software from OMICSHARE platform (https: //www.omicshare.com/), respectively. The protein interactive networks were constructed by the STRING online tool (https://string-db.org/).

RNA-Seq Reliability Verification
To assess the accuracy of sequencing data, we randomly selected 8 banana DEGs and the banana actin gene (an internal control), to do the following Quantitative Real-Time PCR analysis (qRT-PCR). Gene-specific primers (Table S16) were designed and analyzed on the online primer tool (https://www.ncbi.nlm.nih.gov/tools/primer-blast/). RT reagent Kit No. RR047Q (TaKaRa, Beijing, China) were used to reverse transcribe RNA into cDNA. The reaction solution of qRT-PCR was configured by the reagent No. RR820L (TaKaRa, Beijing, China). The reaction equipment is ABI 7900 HT Fast Real-Time PCR System (Applied Biosystems, Foster City, CA, USA). The relative gene expression value was analyzed using the 2 −∆∆Ct formula, and three biological replicates were used in this research. The data was statistically analyzed and figured by Excel.

Conclusions
In this study, the morphological changes and DEGs of banana roots under different potassium stress were systematically understood. The results suggested that different potassium concentrations had various effects on the root phenotype of banana. 4454 DEGs were obtained in roots. Gene function classification enrichment was shown in the paper. six significant expression profiles were conducted. The core genes in each expression pattern were screened out. In conclusion, this study revealed the influence of potassium level on banana root system and the molecular regulation mechanism and provided a new idea for improving the use efficiency of banana potassium.

Supplementary Materials:
The following are available online at http://www.mdpi.com/2223-7747/9/1/11/s1, Figure S1: Interaction network of DEGs in profile 2, Figure S2: Interaction network of DEGs in profile 13. Table S1: Transcript abundance measurements at each potassium concentration treatment, Table S2: GO molecular function analysis of DEGs detected in banana roots in response to different concentration of potassium, Table S3: annotation of genes in profile 0, Table S4: The interaction data of DEGs in profile 0, Table S5: annotation of genes in profile 1, Table S6: The interaction data of DEGs in profile 1, Table S7: annotation of genes in profile 7, Table S8: The interaction data of DEGs in profile 7, Table S9: annotation of genes in profile 9; Table S10: The interaction data of DEGs in profile 9, Table S11: annotation of genes in profile 2, Table S12: The interaction data of DEGs in profile 2, Table S13: annotation of genes in profile 13, Table S14: The interaction data of DEGs in profile 13, Table S15: Annotation of DEGs of the significant GO term in profile 1, Table S16: Primer information for qRT-PCR.