Transcriptome changes induced by drought stress in Jerusalem artichoke (Helianthus tuberosus L.) and weighted gene co-expression network analysis (WGCNA)

Background: Jerusalem artichoke (Helianthus tuberosus L.) is strongly resistant to stress and an important plant used for ecological management in northern China in recent years. Currently, Jerusalem artichoke has been widely planted in the area around Qinghai Lake in Qinghai Province, China. Jerusalem artichoke can not only prevent land desertication but also has maintain most of its level of production. However, there is little research on the mechanism of drought resistance of Jerusalem artichoke. Results: We conducted transcriptome sequencing under drought stress and normal watering treatment for two varieties, QY1 and QY3, with differing degrees of drought tolerance. In the three stress periods of QY1 and QY3, 5,613, 12,985 and 24,923 differentially expressed genes (DEGs) were identied, respectively. GO analysis showed that there were more DEGs in QY1 than in QY3, but there were more up-regulated genes in QY3 than in QY1. Based on an additional analysis of the metabolic pathways under drought stress using MapMan, the most different types of metabolism included secondary metabolism, light reaction metabolism and cell wall. The up-regulated genes in QY3 were signicantly more prevalent than those in QY1 and were primarily concentrated in avor IDS, phenylpropanoids, and the shikimate and terpenoids pathway. As a whole, QY1 and QY3 both had a large number of up-regulated genes in the avor pathway. In addition, the gene analysis of the ABA key metabolic pathway showed that QY3 had more genes in NAC and WRKY than QY1. A weighted gene co-expression network was constructed and divided into modules. By specically analyzing the expressed modules, four modules were found to have the highest correlation with drought. Further research on the genes revealed that all 16 genes related to histone, ABA and protein kinase were the most signicant in these pathways. Conclusions: In summary, these ndings represent the rst RNA-Seq analysis of drought stress in Jerusalem artichoke, which is of substantial signicance to explore the function of drought resistance in Jerusalem artichoke and the unearthing of related genes.


Background
Drought is a major and direct environmental stress factor that threatens plant health, damages the structure and composition of ecological communities and leads to plant death. It usually leads to signi cant economic losses in agriculture and forestry. Current climate change, environmental degradation and soil salinization aggravate the frequency and severity of environmental drought [1]. The increasing severity of soil drought has led to changes in plant gene expression and cell metabolism, posing a growing threat to grassland and animal husbandry. One third of the land area of the world is arid and semiarid, which has a substantial impact on plant production and economic development.
Therefore, improving the drought resistance of plants has become a hot topic of research [2]. Owing to the constant changes of a growing environment, plants are often under different environmental stresses.
Under stress, plants will respond to and adapt to stress through changes in gene expression, metabolites and physiology. Therefore, there should be a variety of stress receptors in plant cells that produce and transduce signals [3]. However, it is possible that the sensing of stress signals is a redundant system, in which the mutations of a single gene or protein will have its function ful lled by other genes, resulting in a lack of obvious phenotype in stress signal sensing. Moreover, it could be that these genes are necessary for the growth of plants, and the functionally de cient mutants lead to the failure of normal growth and subsequent study. Therefore, only a few plant signal receptors have been identi ed. For example, the ABA signaling pathway is a widely studied drought response signaling pathway in plants that is considered to be the "core" signaling pathway under di cult levels of stress [4]. Under drought stress, ABA can regulate stomatal closure, inhibit growth and promote leaf senescence, thus improving the rate of survival of plant. When the plants were subjected to drought stress, their ABA levels increased under the control of SnRK2s (sucrose non-fermenting SNF1-related protein kinase 2s) [5]. ABA activates ABFs and RAV1 through SnRK2s and binds the ABRE (ABA-responsive element-binding proteins) or RAV1-binding ciselement in the promoter region of the NAC transcription factor gene, thus promoting the expression of ORE1, AtNAP and other genes, thereby controlling leaf senescence and enhancing the drought resistance of plants [6,7]. Qinghai Province in China is located on the Qinghai-Tibet Plateau, the roof of the world, with a fragile ecological environment, high altitude and a severe climate. Once the local vegetation has been destroyed, it is di cult for the plants to recover. With the in uence of grazing in the surrounding areas and other factors, sandstorm activities north and east of Qinghai Lake have gradually increased.
Land deserti cation seriously restricts the production of agriculture and animal husbandry in the surrounding arid areas, and the implementation of sandy vegetation has become an important part of ecological construction in the Qinghai Plateau. However, owing to the low amount of precipitation and large evaporation in the plateau area, water has become the most important ecological constraint factor for the transformation of sandy land. In addition, the primitive grazing industry also seriously damages the grassland ecosystem. Therefore, during the process of the restoration of vegetation and construction of grassland ecological sandy land, it is necessary to select ecologically economical plants that can restore the environmental ecology and increase income of farmers and ranchers [8][9][10]. Jerusalem artichoke (Helianthus tuberosus L.) is a type of plant that is highly adaptable to infertile lands, drought, cold and salt and is suitable for growing in dry, cold, sunny climates and soil with a high concentration of sediment [11]. Jerusalem artichoke tubers are rich in inulin, which comprises 70-90% of their dry matter.
Inulin is a polymer of fructose. It can be decomposed into a single fructose by exo-inulinase and then fermented to produce ethanol with a conversion rate of up to 83-99%. Jerusalem artichoke has the advantages of a low planting cost and high biological yield. Therefore, Jerusalem artichoke will be an important plant for fuel ethanol and the large-scale development of the inulin industry with substantial strategic signi cance and broad prospects for application [12,13]. In recent years, the planting area of Jerusalem artichokes has been increasing yearly, particularly in the deserti cation and saline-alkali areas of north China. However, studies on the drought stress of Jerusalem artichoke are primarily only focused on the selection of varieties, physiological responses, sugar metabolism and water use e ciency [14][15][16][17] in the tropics. Because Jerusalem artichoke is a hexaploid species (2n = 6x = 102), and its genetic background is highly heterozygous and complex [18], the study of its mechanism of drought resistance has lagged behind that of other crops. To our knowledge, no comprehensive study on the drought adaptation mechanisms of Jerusalem artichoke has been reported. However, the study of the mechanism of the drought adaptation of Jerusalem artichoke is a prerequisite for its increased utilization in arid areas and also an important basis for its cultivation and management. Currently, Jerusalem artichoke has been widely planted in the area surrounding Qinghai Lake in Qinghai Province, China. Jerusalem artichoke can not only prevent land deserti cation but also has loses little production. It is valuable in edible and industrial development, can create a long industrial development chain, and is of substantial signi cance to the improvement of regional ecological environment and economic development. This study explores the drought-resistant physiological mechanism of Jerusalem artichoke by simulating the drought stress of Jerusalem artichoke in the desert environment. This experiment simulated natural drought to cause stress to two genotypes of Jerusalem artichoke at the seedling stage, and the season in which Jerusalem artichoke was planted in Qinghai was selected for the experiment. The seeds were planted on April 28, and the Jerusalem artichoke plant was approximately 25 cm high in early June. Jerusalem artichoke with the same height and similar growth levels were selected for the experiment, and samples were collected from June 4 to June 10. Figure 1 indicates that, compared with the water treatment (WT), there was no signi cant change in both genotypes of Jerusalem artichoke plants under drought treatment (DT) at 2 d. QY1 wilted at 4 d, and most of the leaves of the plant withered at 10 d, while QY3 showed no signi cant wilting at 4 d, and most of the leaves remained normal at 10 d. Moreover, it can be seen from the color changes at the root that at 2 d of stress, QY1 became darker, while the color of QY3 did not become signi cantly darker until 10 d (Fig 1).
Fig 2a indicates that with the development of drought stress, the water content of DT group decreased to less than 10% in 10 days, while that of the WT group remained basically stable at 35%, and the water content of QY3 decreased faster than that of QY1. In contrast, Fig 2b shows that the content of dry matter in roots of two genotypes of Jerusalem artichoke increased signi cantly under drought stress. On the 2nd and 4th days after the plants were subjected to drought stress, there was no signi cant difference between the WT and DT groups. However, on the 10th day, the DT group reached its highest value. The difference between the WT and DT groups became highly signi cant, and the difference between the two genotypes of Jerusalem artichoke in DT group also became highly signi cant, with QY3 exhibiting higher levels of drought stress than QY1. The rate of accumulation of dry matter by the DT group was higher than that of the WT group. Next, we measured the contents of protein and chlorophyll. With an increase in treatment time, there was no signi cant difference in the contents of protein between DT and WT groups, but the DT group exhibited a trend of higher protein contents than those of the WT group. There were also signi cant differences between the two genotypes of Jerusalem artichoke in DT group on the 2nd and 10th days, with QY3 exhibiting higher trends than QY1. There was no signi cant difference between the WT and DT groups in their level of chlorophyll content on the 2nd and 4th days, but there was a very signi cant difference between these groups on the 10th day. QY1 and QY3 exhibited the same type of trend under drought stress. However, the range of change in trend of QY3 was larger than that of QY1 (Fig  2c and d). The results showed that continuous drought stress has a substantial effect on the growth of Jerusalem artichoke.

Characterization of RNA-Seq
The leaves of two Jerusalem artichoke varieties were sequenced under normal water and drought conditions. A total of 89,550,751 raw reads were ampli ed. The sequenced clean reads and clean bases were 89,551,965 and 13,508,637,885, respectively. The average values of Q20 and Q30 were 97.66% and 93.265%, respectively, indicating that the sequencing quality was reliable. After data ltering and quality control, 89,551,965 clean reads were produced with an average of 3,198,284 clean reads per sequencing library and an average GC content of 45.585%. After assembly with Trinity, 309,323 unigenes and 455,817 transcripts were obtained. The average GC content of the unigenes and transcripts was 42.24% and 41.48%, respectively; the average N50 of the unigenes and transcripts was 727 bp and 788 bp, and the mean length of the unigenes and transcripts was 571 bp and 619 bp, respectively (Table 1). After splicing and assembling the clean reads, 455,817 transcripts were obtained with an average length of 619 bp and N50 of 788 bp. The number of transcripts between 200 and 400 bp was the most frequent at 200,512, accounting for 18.3% of the total (Fig 3).
The differentially expressed genes (DEGs) between the two Jerusalem artichoke varieties were screened using the DESeq method for subsequent gene function annotation and screening. The results of the difference data are shown in Figure 3. Under drought stress, genes in the QY1 and QY3 leaves diverged signi cantly. At the second day of drought stress, there were 5,613 genes differentially expressed in QY1/QY3, of which 4,091 genes were up-regulated, and 1,522 genes were down-regulated. At the fourth day of drought stress, there were 12,985 genes differentially expressed in QY1/QY3, of which 6,346 genes were up-regulated, and 6,639 genes were down-regulated. At the tenth day of drought stress, there were 24,923 genes differentially expressed in QY1/QY3, of which 12,081 genes were up-regulated, and 12,842 genes were down-regulated. With the increase in the trend of stress time, the number of up-regulated and down-regulated unigenes both increased, indicating that drought stress inhibited the expression of some unigenes and induced the expression of a number of other unigenes (Fig 4).
Fragments per kilobase of transcript per million (FPKM) > 0.3 was used as the standard pf expression to calculate the gene expression of the two Jerusalem artichoke varieties under normal water and drought stress, as shown in Fig 5. Under WT treatment, 39,569 unigenes were screened out in QY1, including 16,972 nonshared unigenes. A total of 3,635 unigenes were shared between two groups, 1,865 unigenes among three groups and 2,433 unigenes among four groups. In QY1, 9,601, 9,374, 8,632 and 11,962 unigenes were screened at 0 d, 2 d, 4 d and 10 d, respectively. The overall trend of expression of QY1 under the DT treatment was similar to that of WT group, but the number of unigenes increased signi cantly. A total of 43,650 unigenes were screened out, including 18,866 nonshared unigenes, 3,927 shared unigenes between two groups, 2,066 shared unigenes among three groups and 2,683 shared unigenes among four groups. In QY1, 9,601, 9,462, 11,132 and 13,485 unigenes were screened at 0 d, 2 d, 4 d and 10 d, respectively. More genes were expressed in the DT group than the WT. In QY3, there were more unigenes screened out under the WT and DT treatments than those in QY1. Among them, there were 44,666 unigenes under WT, including 19,091 in nonshared unigenes, 4,129 shared unigenes between two groups, 2,191 shared unigenes among three groups and 2,686 shared unigenes among four groups. In QY3, 13,031, 9,723, 11,084 and 10,828 unigenes were screened out at 0 d, 2 d, 4 d and 10 d, respectively.
According to the unigene detection results under drought stress, GO function classi cation was conducted for the DEGs. Since the difference was the most obvious at 10 d, the DEG of the two varieties at 10 d of the WT and DT treatments were selected for further functional analysis. After 10 d of drought stress in QY1, 9,394 DEGs were annotated with 43 functions, among which 19 belonged to biological process, 13 to cellular component and 11 to molecular function. After 10 d of drought stress in QY3, 10,796 DEGs were annotated with 41 functions, among which 18 belonged to biological process, 13 to cellular component and 11 to molecular function. The three pathways with the highest differential expression in biological process, cellular component and molecular function were analyzed. In biological process, the differences primarily existed in the metabolic process, cellular process and single organism process, with the DEGs in QY1 comprised of 561, 493 and 451 more than those in QY3. A similar situation also occurred in cellular component and mobile function. In the cellular component, most of the different pathways were membrane, cell part and cell, with the DEGs in QY1 numbering 294, 451 and 268 more than those in QY3, respectively. In molecular function, QY1 had 553, 299 and 64 more DEGs in the quantitative activity, binding and transporter activity channels, respectively. QY1 had more DEGs than QY3 in most pathways, but in general, there were more up-regulated genes than down-regulated genes in QY3. According to the data of differential expression, drought stress has a signi cant impact on the biological functions described above (Fig 6).
To further understand the metabolic pathway and other processes of drought stress genes in Jerusalem artichoke, MapMan was used to compare the metabolic pathways of differential genes of QY1 and QY3 under the DT and WT treatments based on the results of RNA-Seq. A visual analysis found that there were signi cant differences in the expression of drought-sensitive genes in the metabolic pathways. As shown in Figure 7, the metabolic pathway was divided into minor CHO metabolism, cell wall metabolism, lipids metabolism, secondary metabolism, amino acids metabolism, light reaction metabolism, carbohydrate metabolism, nucleus metabolism and tetrapyrrole metabolism. Through the comparison of the QY1 and QY3 DT/WT, we can observe that there was a large difference in the overall metabolic pathway between QY1 and QY3. The most different metabolic pathways were secondary metabolism, light reaction metabolism and cell wall. Compared with QY1 (Fig 8a.), the up-regulated genes of QY3 were primarily cell wall, secondary metabolism and light reaction metabolism. The number of up-regulated genes was 130, 98 and 83, respectively, and the down-regulated genes were also primarily concentrated in these three pathways. Secondary metabolism helps plants to resist environmental stress, improve their selfpreservation and ability to compete and maintain their survival and development. Therefore, the study of this area is highly signi cant for the study of the secondary metabolic accumulation of industrial cash crops and their production. Drought, as one of the primary environmental stresses, affects the accumulation of secondary metabolites. We used the secondary metabolic pathway with the largest number of differential genes for further mining. In Fig 7c and d, we continue to use red for up-regulation and blue for down-regulation. The up-regulated genes in QY3 were signi cantly more abundant than those in QY1 and were primarily concentrated in avonoids, phenylpropanoids, and the shikimate and terpenoids pathway. The results described above show that under the stress of continuous drought, two types of Jerusalem artichoke exhibited different drought resistance mechanisms, and the synthesis of cell wall, lipid, photosynthesis and secondary metabolism of QY3 were signi cantly higher than that of QY1. The primary component of the plant cell wall is cellulose, and the primary component of the intercellular layer of cell wall is pectin. Waxy synthesis, cellulose synthesis and the up-regulated expression of pectin methylesterase all indicated the enhancement of cell wall synthesis. QY1 has more up-regulated and down-regulated genes in carbohydrate metabolism than does QY3.
Endogenous hormones, as response factors to plant stress, play an important role in the adaptation to drought stress. ABA, as one of the primary endogenous hormones, plays an important role in plant stressresistant growth, particularly under drought stress, and is even considered to be a potential indicator of crop drought resistance. In this study, 14 ABA signaling, 21 NAC, 7 DREB and 24 WRKY genes were selected for expression analysis. Some of them were both up-regulated in QY1 and QY3, including HAB2 (PP2C member of ABA signaling), ABA responsive element binding protein AREB1, NAC29 (NAC family), DREB2a (DREB family) and eight WRKY transcription factors: WRKY13, WRKY5, WRKY7, WRKY9, WRKY42, WRKY4, WRKY68, and WRKY8 (Fig 8.). NAC1 and some WRKY genes were down-regulated in both varieties of Jerusalem artichoke. On the whole, QY3 showed more variation in the NAC and WRKY genes than did QY1, particularly in the NAC family. In addition, QY3 showed an obvious trend of change in three SNRK genes, but the number of genes responding to drought stress was signi cantly higher in QY1.
To better understand the drought-tolerant gene expression regulatory network of Jerusalem artichoke, weighted gene co-expression network analysis (WGCNA) was used to analyze the drought-related genes. Based on the results of high mean connectivity and low scale independence, we nally chose β=14 as the threshold. The analytical results showed that 6,643 drought-related genes of QY1 could be clustered into 27 modules under drought stress, while 7,224 drought-related genes in QY3 could be clustered into 28 modules (Fig 9.). The WGCNA analysis showed that under drought stress, re-writing and de-writing existed in the co-expression network of a large number of genes of QY1 and QY3. We selected the dark slate blue (r = 0.45, p = 0.02), dark red (r = 0.38, p = 0.04), ivory (r = 0.42, p = 0.03) and brown (r = -0.39, p = 0.04) modules with the highest correlation with stress to perform a separate comparative analysis. The dark slate blue module had a signi cant correlation of expression in variety, stress and period traits (Fig  9d.). In the module of the trend of expression diagram that was drawn, QY1 and QY3 showed a consistent trend of expression in the dark slate blue module under the WT, but they showed an opposite trend of expression under the DT treatment. In addition, QY1 and QY3 had the same trend of expression in the ivory, dark red and brown modules in the DT group.
The functions of the genes in four stress resistance-related co-expression modules were analyzed. The GO enrichment results showed that 319 genes in the brown module were enriched in the items of molecular function (GO:0003674), catalytic activity (GO:0003824) and metabolic process (GO:0008152). A total of 130 genes in the dark red module were enriched in the items of cell component (go: 0005575), cell (go: 0005623) and cell part (go: 0044464). Four genes in the dark slate blue module were enriched in transferase activity (go: 0016740), and 24 genes in the ivory module were enriched in response to hormone stimulus (go: 0009725), response to endogenous stimulus (go: 0009719), response to organic substance (go: 0010033) and response to chemical stimulus (go: 0042221). In general, the genes enriched from the brown and dark red modules had the greatest number of molecular functions, while the primary GO items enriched from brown and ivory modules did not contain a molecular function (Fig 10.). To further explore the candidate genes involved in the drought stress of Jerusalem artichoke, the candidate genes with the highest correlation with drought stress were screened out from the GO classi cation of four modules. Through the function of gene annotation, it was found that the 16 genes selected by annotation belong to histone, ABA regulation and protein kinase among others. From the annotated species, there were eight candidate genes annotated to Cynara cardunculus, indicating that the mechanism of drought stress of Jerusalem artichoke is similar to that of C. cardunculus (Table 2).

Discussion
Plant growth and development are affected by various external environmental factors. Drought stress is one such factor. The comprehensive analysis of the transcriptome helps to understand the expression and regulatory mechanism of genes in plant systems. Currently, the application of transcriptome technology to the research of multiple species has advanced substantially [19][20][21][22]. In this study, RNA-Seq was used to conduct a transcriptome analysis on two Jerusalem artichoke varieties after drought stress treatment and compare the changes of gene expression in plants after drought stress treatment and a control of water treatment. We obtained important drought-stress response genes by comparing 0 d, 2 d, 4 d and 10 d drought-stress treatment with the watered control. These results will help to understand the regulatory mechanism of Jerusalem artichoke seedlings under drought stress.
The Jerusalem artichoke is a highly heterozygous hexaploid crop and basically reproduces asexually using tubers. In addition, due to the long term of introduction, many planting resources with similar genetic background lead to the single planting resources of Jerusalem artichoke [23]. Although Jerusalem artichoke did not originate in an arid area, and the species has not undergone long-term selective evolution, the tuber of Jerusalem artichoke is rich in fructan, which can greatly improve the resistance of crops. It is reported that drought and high temperature stress will cause a loss of 29% of the tubers and a loss of 53% of biomass in tropics [16]. Studies that have been conducted on the stress resistance of Jerusalem artichoke are mostly related to salt stress. Under salt stress, the increase in the metabolism of proline in Jerusalem artichoke promotes osmotic regulation and protection [24]. However, the accumulation of underground dry matter and e ciency of transport of non-reducing sugar to tubers continued to decrease, and the synthesis of IAA and GA3 in the leaves and tubers was inhibited, while the synthesis of ABA signi cantly improved [25]. A comparison between the northern variety of QY2 and the southern variety of NY1 indicated that the ABA content in the tuber of QY2 was higher [26]. The microbial community in the root of the crops could also play an important role under salt stress [27,28]. With the development of research on the abiotic stress of Jerusalem artichoke, the stress resistance of Jerusalem artichoke has been widely recognized. The goal of this research on the drought resistance of Jerusalem artichoke is to clarify the mechanism of the response of Jerusalem artichoke to drought. The most economical and effective way to solve the problem of drought is to select and cultivate new droughtresistant varieties. Current studies have found that drought stress can affect the physiological and biochemical characteristics of Jerusalem artichoke. Varieties differ in their susceptibility to drought. Drought can reduce the dry weight of the tuber and biomass of Jerusalem artichoke, but the water use e ciency of the yield of tubers and the water use e ciency of biomass are enhanced [14,16].

Effects of drought stress on the metabolic pathway of avonoids in Jerusalem artichoke
In this experiment, " avor biosynthesis" was only signi cantly enriched in QY3 under severe drought stress, indicating that it participated in the drought stress response and was more active in the drought resistant variety QY3. In fact, the relationship between avonoids and drought tolerance has gradually been con rmed. Flavonoids have a strong ability to scavenge free radicals, which helps to alleviate the oxidation reaction of Arabidopsis thaliana and its adverse reactions under drought stress [29]. Flavonoids are secondary metabolites and can remove the singlet oxygen produced by environmental stress in plants; they can remove the reactive oxygen produced by environmental stress and prevent the oxidative damage of reactive oxygen species to plants, and they play important roles in various abiotic stresses [30,31]. Drought stress can affect the expression of the genes of related enzymes in the pathway of avonoid metabolism in plants, but this differential expression varies among plants. For example, in Antirrhinum, the expression of phenylalanine ammonia lyase, chalcone isomerase, dihydro avonol reductase, pyrroline-5-carboxylate synthase, superoxide dismutase and peroxidase was signi cantly upregulated under drought stress [32], while in grape, the avonoid content in the leaves decreased signi cantly after drought stress [33]. The up-regulated or down-regulated expression of genes for related enzymes can promote or inhibit the synthesis of the corresponding avonoids, which is a part of the selfregulatory mechanism of plants in response to drought stress. In this study, in addition to the signi cant enrichment of avonoids in QY3, MapMan analysis showed that the number of up-regulated genes in the anthocyanin biosynthesis of QY3 was greater than that in QY1. As a whole, QY1 and QY3 both have a large number of up-regulated genes in the avonoid pathway. It has been hypothesized that avonoids could help Jerusalem artichoke resist drought stress and maintain normal metabolic activities. However, this study only analyzed the whole avonoid-related gene expression. Therefore, the relationship between avonoids, abiotic stress and the speci c metabolic pathway changes in Jerusalem artichoke merits further study.

Transcription factors and plant drought tolerance
Drought stress can lead to changes in the content and activity of endogenous hormones in plants, thus, regulating the physiological and biochemical processes of plants, as well as the transport and distribution of assimilates. ABA, as one of the primary endogenous hormones, plays an important role in plant growth, particularly under drought stress, and is even considered to be a potential indicator to measure crop drought resistance [34]. Transcription factors known to be associated with drought response are primarily composed of the MYB, MYC, DREB, AP2/EREBP, bZIP and NAC gene families. In this study, an analysis was conducted on the expression of the gene families that were prone to change under drought stress. Among them, ABREs in the bZIP family-controlled gene expression through the AREB and ABF transcription factors. It was up-regulated in response to drought stress and could interact with SRK2D/SnRK2.2 to enhance the expression of ABA [35]. As early as 2005, Yasunari et al. found that the overexpression of AREB1 in A. thaliana can regulate the ABA signaling pathway, and thus, enhance the drought tolerance of the plants [36]. Current research shows that a ChIP and electrophoretic mobility shift assay con rmed that AREB1 can combine with the ABRE promoter motif of the NAC gene and activate PtrNAC006 related to drought tolerance, PtrNAC007 and PtrNAC120 genes [37]. Although we did not study the regulation of AREB in this study, the expression of AREB1 in QY1 and QY3 increased with the increase of drought stress time, particularly in QY3, which may indirectly improve the drought tolerance of QY3. In addition, we also analyzed the expression of the NAC family and found that NAC29 had almost the same trend of expression as AREB1. Similarly, the NAC family gene was signi cantly correlated with the drought tolerance of plants. In maize, the level of expression of the ZmNAC55 gene was signi cantly up-regulated after drought stress, and the promoter region of this gene had multiple stress-related cis-elements [38]. The overexpression of GmNAC085 in soybean increased the activity of superoxide dismutase, catalase and ascorbic acid peroxidase, thus reducing the oxidative damage caused by drought [39]. The NAC transcription factor JUB1 in tomato is the drought-tolerant regulator of tomato. SlJUB1 controls the expression of SlDREB1, SlDREB2 and SlDLALA, while AtJUB1 in A. thaliana stimulates the expression of DREB2a and produces drought resistance by controlling the DREB genes [40,41]. DREB2 and DREB2a were also highly expressed in both varieties of Jerusalem artichoke, particularly in QY3, in which the expression of DREB2 was extremely high. It has been proven that DREB2 was closely related to drought stress in wheat [42], A. thaliana [43] and mustard [44]. Compared with the DREB family, which had only a few genes, we identi ed a number of WRKY transcription factors in the Jerusalem artichoke transcriptome data. The high expression of WRKY primarily appeared in the QY1. Meanwhile, eight WRKY transcription factors were up-regulated in both varieties of Jerusalem artichoke. WRKY42 and WRKY4 showed the strongest responses in QY1 and QY3, respectively, suggesting that these two WRKY transcription factors may play an important role in the response of Jerusalem artichoke to drought stress. These ndings also provide a theoretical basis for further studies on the role of transcription factors in the regulation of plant responses to environmental stress and provide data to support the future screening of drought-tolerant varieties of Jerusalem artichoke.

WGCNA analysis
WGCNA is an effective method for co-expression network analysis, which can speci cally screen out coexpression modules with high biological signi cance to target traits. It has been proved to be an e cient data mining method in rice [45], Brassica rapa [46], and sun ower [47]. In this study, the WGCNA method was used to construct a weighted gene co-expression network based on the sequencing data of 28 drought stress transcriptomes of Jerusalem artichoke. Four modules related to drought stress were excavated, among which the dark slate blue and brown modules were positively correlated and negatively correlated with drought stress. Through GO enrichment analysis of the genes in modules, the coexpressed genes were found to be associated with important activities, such as molecular function, cellular component, transferase activity and response to hormone stimulus. Key genes in these pathways were screened to obtain 16 genes with the highest degree of enrichment, most of which were related to histone and ABA regulation and protein kinases. There were four genes that were the most signi cantly related to histone in the dark red module. Research shows that histone plays an important role in plant biological and abiotic stress. The basic unit of the nucleosome of chromatin is comprised of a 146 bp sequence of DNA and a wrapped histone octamer by the DNA (two copies of H2A, H2B, H3 and H4). Chromatin tissue is easily regulated during the short term of gene expression, involving signal pathways of plant hormones, such as ABA, ethylene, jasmonic acid and brassinosteroids, which are mediators of environmental stress recognition and transcriptional factor activation [48]. In addition, Aux/IAA-ARFdimerization in the ivory module and Protein phosphatase 2C (PP2C), Alpha tubulin and the Seven-inabsentia protein (SINA) in the brown module are also closely related to drought. Under drought stress, alpha tubulin can change the characteristics of the cell walls to resist adverse plant reactions [49].
Studies have shown that the Aux/IAA gene may be regulated by light, plant hormones and abiotic stress, and drought can signi cantly reduce the expression of Aux/IAA4 [50]. There are many studies on PP2C, and PP2C may play a role in maize signal transduction in response to drought stress [51]. PP2C acts as a negative regulator of the ABA signal and can dephosphorize SAPK8 and SAPK10 in vitro, and together with PYL10, SAPK8 AND SAPK10, it can form an ABA signal transduction module that is involved in the response to stress [52] [53]. SINA is the link gene between drought resistance and ABA signaling [54]. In A. thaliana, ABA and drought treatment signi cantly induced SINA, and the overexpression of SINA increased the closure of stomatal and reduced water loss, thus, improving the drought resistance of transgenic plants [55]. The important ABA regulatory genes, such as AREB1, AREB2 AND ABF3, are all induced by SINA [56] [57]. The results described above show that WGCNA can divide complex data into co-expression modules, re ect the effective information in the real biological network, better mine the core genes related to the drought resistance of Jerusalem artichoke, predict the function of potential candidate genes and provide new research ideas for the research of the gene regulatory network of different plant characteristics.

Conclusions
In this study, a comparative transcriptome analysis of the DEGs among two varieties of Jerusalem artichoke with differing degrees of response to drought tolerance was performed. A gene ontology (GO) analysis showed that there were more DEGs in QY1 than in QY3, but there were more up-regulated genes in QY3 than in QY1. The MapMan results suggested that avonoids could help Jerusalem artichoke resist drought stress and maintain normal metabolic activities, and the ABA key metabolic pathway responds to more up-regulated genes in QY3. Further research on the genes revealed that all 16 were related to histone, ABA and protein kinase were the most signi cant in WGCNA module, which provide new insights for the function of drought resistance in Jerusalem artichoke and the ability to unearth related genes.

Plant materials and drought treatments
Two Jerusalem artichoke varieties "QY1" and "QY3" were selected in our laboratory and were drought sensitive and drought resistant, respectively. The experiment was conducted at the experimental base of the Institute of Horticulture, Qinghai Academy of Agricultural and Forestry Sciences, Chengbei District, Xining City, Qinghai Province (N36°43' 45.21'' E101°44' 57.69''). The experimental tubers were planted in a plastic culture bowl lled with sand in April. One culture bowl represented one biological repeat. Each bowl was placed into the incubator for cultivation (14 h day 22ºC/10 h night 18℃) and continued normal watering. Two water treatments were established in the experiment. For the drought treatment, each culture bowl was watered evenly and weighed to 2 kg, and then water was no longer supplied. Samples were collected at 2 d, 4 d and 10 d. For the normal water treatment, each culture bowl was watered evenly and weighed to 2 kg. The drought and normal water treatments were both established with two repetitions.

Phenotypic data determination
The contents of soil water, root water, soluble protein and chlorophyll were determined under both treatments. The formula to determine the soil moisture content was SWC=(SFW-SW)/SFW, in which SFW was the soil fresh weight, and SW was the soil weight. The formula to determine the root water content was RWC=(RFW-RW)/RFW, where RFW was the root fresh weight, and RW was the root weight. The protein content was determined by the Coomassie blue G250 method. Half a gram of Jerusalem artichoke leaves was sampled and ground with a small amount of quartz sand and 10 ml LPBS for extraction. After grinding, the leaves were centrifuged twice (12,000 g for 6 min and 20,000 g for 16 min). The supernatant was utilized as the protein extract. A volume of 0.1 ml of the protein extract was mixed with 5 ml Coomassie blue reagent, and its absorbance was measured at a wavelength of 595 nm after 2 min. The protein content of the plant materials was calculated using a standard curve. Determination of chlorophyll content: The Jerusalem artichoke leaves were removed to clean the dirt off the surfaces, cut into pieces with the midribs removed and mixed well. A total of 0.2 g of three fresh cut samples were weighed and placed into mortars. A small amount of quartz sand, calcium carbonate powder and 2-3 ml of 95% ethanol was added, and the sample was ground into a homogenous pulp. A volume of 10 ml of ethanol was added and continuously ground until the tissue turned white. The sample was incubated at room temperature for 3-5 min. A piece of lter paper was placed in a funnel and moistened with ethanol.
The extract was poured into the funnel along a glass rod and ltered into a 25 ml brown volumetric ask. The mortar, rod and residue were washed several times with a small amount of ethanol, which was nally poured into the funnel together with the residue. The chloroplast pigment extract was poured into a colorimetric cup with an optical path of 1 cm. The absorbance was determined at wavelengths of 665 nm and 649 nm with 95% ethanol as a blank. The chlorophyll content was calculated using the following formula: RNA Preparation, cDNA library preparation and Illumina sequencing.
An Omega E.Z.N.A Total RNA Kit was used to extract the total RNA from the samples, and the purity and integrity of the extracted RNA samples were determined using 1.0% agarose gel electrophoresis.
Subsequent sequencing libraries were constructed for the RNA samples to conform to 1.8 < OD260/280 < 2.0, OD260/230 >1.0, RNA concentration > 250 ng/μl and RIN > 8.0. After extracting the total RNA of the sample, the eukaryotic mRNA was enriched with magnetic beads with Oligo (dT). Fragmentation buffer was added to break the mRNA into short fragments. The rst cDNA strand was synthesized by random hexamers using mRNA as a template, and the second strand was synthesized by adding buffer, dNTPs, RNase H, and DNA polymerase I. After puri cation with a QiaQuick PCR kit and elution with EB buffer, the synthesized RNA was repaired at the end. Poly (A) was added with a sequencing connector. The RNA was then subjected to agarose gel electrophoresis for size selection and nally ampli ed. The constructed sequencing library was sequenced with an Illumina Hi-SeqTM 2500 platform. Raw reads generated by sequencing were rst screened and controlled for data quality. The reads with connector sequences were removed, as were the reads containing poly-N and the low-quality reads produced by sequencing. Trinity software was used for sequence assembly [58], and clean reads were nally obtained for further analysis.
The contents of Q20, Q30 and GC of the clean reads were also calculated. Gene co-expression network analysis The WGCNA method was used to construct a gene co-expression network for all the drought-related genes of the two different genotypes of Jerusalem artichoke. Co-expression analysis was conducted on 12 samples treated under drought stress and 16 samples treated with normal water supplementation.
Using the method in the R v. 3.5.1 package of WGCNA, the expression matrix was constructed according to the FPKM expression data of the DEGs of all the samples. The expression matrix was standardized using a Z-score, and each gene was analyzed by cluster analysis. The genes of different expression types were classi ed into modules through dynamic tree cut, and the minimum number of genes for each coexpression module was 50 [62]. Some similar modules were merged, and the correlation between different modules was calculated using 0.25 as the similarity threshold. Three traits were set as variety, stress and period, respectively. The correlations between the modules and variety and stress and period were calculated through correlation analysis. GO enrichment analysis was performed on the stressrelated co-expression modules, and signi cant genes in the major enrichment pathways were selected from the enrichment results for functional prediction analysis.     Expression heatmap of abscisic acid key metabolic pathway-related genes in QY1 and QY3

Figure 9
Analysis of the co-expression network of drought stress-related genes. Weighted gene co-expression network analysis (WGCNA) was used to analyze the co-expression network of drought stress-related genes, and the genes were clustered into different co-expression modules.