2.4.4. Module Visualize and Hub Genes

We performed genes trend expression analysis of four modules of interest (Figure 12). Interestingly, by analyzing the trend expression of genes in sienna3 module, we found that they are specifically expressed only in the stem, which suggests that they play an important role in the stem. Through further analysis of this module, several hub genes were discovered, such as beta-glucosidase and granule-bound starch synthase (*GBSSI*). This indicates that the specific expression of these polysaccharide-related genes in stems may be related to the content of polysaccharides in stems significantly higher than in other tissues.

In the blue module, Do\_S shows specific expression in this module, several hub genes were detected in each important module, with each gene interacting with many other genes. These genes include bHLH, beta-glucosidase, alpha-L-fucosidase, and other genes related to *Dendrobium* polysaccharide and secondary metabolism. In salmon4 module, Do\_S shows specific expression in this module, hub genes include *CCR*4 (cinnamoyl -CoA reductase), bZIP, and MYB transcription factors were identified. In lightsteelblue1, a large number of genes were annotated to *ketohexokinase* (*KHK*). *KHK*, also known as fructokinase, which is a key enzyme in fructose metabolism. Fructose is a kind of plant polysaccharide, which suggests that this module is related to polysaccharide biosynthesis.

**Figure 12.** Expression profile and transcriptional regulatory network associated with the tissue-specific modules. (**a**), (**c**), (**e**), and (**g**) heatmaps showing genes in module that were significantly over-represented in Dm\_L, Dm\_R, Dm\_S, Do\_L, Do\_R, Do\_S, Dh\_L, Dh\_R, Dh\_S and predicted transcriptional regulatory

network associated with the gene sets showing expression patterns at Dm\_L, Dm\_R, Dm\_S, Do\_L, Do\_R, Do\_S, Dh\_L, Dh\_R, and Dh\_S. Heatmaps show the expression profile of all the co-expressed genes (number given on the top) in the modules (labelled on top). Candidate hub genes are shown in rectangular shapes. Purple in figure (**b**, **d** and **h**) represents genes related to polysaccharides. Red in figure (**d** and **f**) represents genes related to secondary metabolites. The shapes, from big to small, indicate the weights from big to small. Green in figure d represents bHLH transcription factors. Green in figure (**f**) represents bZIP transcription factors and yellow represents MYB transcription factors.

#### **3. Discussion**

Traditional biological research focuses on elucidating the effects of individual functional elements (such as DNA, mRNA, and protein) on life activities at the molecular level. Although those methods are of great significance for revealing the genetic mechanisms of specific traits, it can only partially explain the cause of a certain life activities. With the rapid development of sequencing technology, traditional biological research cannot fully and effectively explore the biological significance contained in massive data. As a research method of systems biology, the network is widely used in the exploration of life sciences with the help of data of genome, transcriptome, and metabolome. Compared with other regulatory networks, WGCNA can screen for genes related to specific traits and perform modular classification from large samples to obtain highly biologically significant co-expression modules, which has proven to be an efficient data mining method [35].

WGCNA has been widely used in plants in recent years. In order to obtain key expression modules and key hub genes related to drought resistance in *Brassica napus* L., WGCNA was used to analyze *Brassica napus* transcriptome data in multiple samples (48 transcriptome data), the well-watered and droughted networks contained 17 and 20 modules, respectively, suggesting that there are additional expression patterns in the droughted network because of rearrangement of the transcriptome in response to the drought treatment. [36]. In the study of *Fragaria* L. flowers [37], researchers generated different tissue- and stage-transcriptomic profiling of woodland strawberry (*Fragaria vesca*) flower development, they discovered a developing receptacle-specific module exhibiting similar molecular features to those of young floral meristems and hub genes of the strawberry homologs of a number of meristem regulators, including LOST MERISTEM and WUSCHEL in the developing receptacle network. [37]. Analysis of the pollen transcriptome of three male sterile lines using weighted gene co-expression network analysis revealed that two modules were significantly associated with male sterility and many hub genes that were differentially expressed in the sterile lines [38]. Farcuh et al. used WGCNA to investigate sugar metabolism during leaf and fruit development of two Japanese plum varieties, and identified 11 key sugar metabolism-related genes, the results showed that sugar metabolism was reprogrammed in a non-climacteric bud mutant of a climacteric plum fruit and showed an increase in sorbitol synthesis [39]. In *Ginkgo biloba*, a total of 12 gene modules were revealed to be involved in flavonoid metabolism structural genes and transcription factors by constructing co-expression networks, they reveal that some hub genes operate during the biosynthesis by identifying transcription factors (TFs) and structure genes and seven key hub genes were also identified by analyzing the correlation between gene expression level and flavonoids content [40]. Through these studies, it was found that many of the hub genes obtained by WGCNA analysis were indeed very important genes. In order to obtain the hub gene related to the synthesis and regulation of Dendrobium polysaccharides and alkaloids, we also performed WGCNA analysis.

Therefore, we constructed a *Dendrobium* gene co-expression network using a WGCNA approach and identified co-expression modules using transcriptome data from three kinds of *Dendrobium* species and three different tissues. Correlation analysis between co-expression modules and two traits (species and tissues) was carried out, and four highly significant active ingredients-related modules (*p*-value < 0.05) were identified. These modules consist of highly connected functional genes, and different modules appear to be involved in individual functions [41]. Meanwhile, KEGG pathway enrichment analysis of modules associated with polysaccharide and secondary metabolism indicated that these pathways in different *Dendrobium* species and different tissues are related to each other at the transcriptomic level. In sienna3 module, we found that several hub genes related to polysaccharide biosynthesis, such as *KHK* (*ketohexokinase*) which is a key enzyme in fructose metabolism. In the expression trend of this module, the stem has a clear advantage, indicating that the polysaccharide content is indeed concentrated on the *Dendrobium* stem, which is consistent with our previous determination of the polysaccharide content. Meanwhile, in blue module, we found that not only the hub genes related to polysaccharide, but also related to secondary metabolism. In the expression trend, we found that the stem of *Dendrobium o*ffi*cinale* to be the highest. This result is consistent with the trend in the determination of polysaccharides and alkaloids. The results also indicated that one component can be regulated by multiple modules, and one module can simultaneously be associated with multiple components.

Module hub genes are generally considered representative of a given module in a biological network. Previous studies reported that MYB-bHLH-WDR (MBW) ternary complexes comprise the essential regulatory machinery for catechin and anthocyanin biosynthesis [42]. In the present study, transcription factors MYB and bHLH were identified as hub genes in modules related to secondary metabolism. In addition, several genes involved in polysaccharide and secondary metabolism biosynthesis (*CCR4: cinnamoyl -CoA reductase* and *KHK: ketohexokinase*) were identified in modules. MYB transcription factors play an important role in the regulation of phenylpropane biosynthesis. Phenylpropane synthesis is upstream of the regulation of flavonoid biosynthesis, indicating that MYB is also an important transcription factor for the synthesis of flavonoids. In gentian, both GtMYBP3 and GtMYBP4 can activate the gene expression of flavonol synthesis, and then significantly increase the flavonol content in seedlings [43]. Ginkgo GbMYBF2 inhibits the expression of *CHS* (*chalcone synthase*), *F3H* (*flavanone 3-hydroxylase)*, *FLS* (*flavonol synthase*, Flavonol synthase and ANS genes) on the phenylpropane synthesis pathway, thereby reducing the content of flavonoids and anthocyanins [44]. *Salvia miltiorrhiza* SmMYB39 affects the synthesis of rosmarinic acid by regulating the expression of key enzyme genes of the phenylpropane metabolic pathway [45]. Current studies indicate that plant bHLH transcription factors are involved in regulating various signal transduction and anabolic pathways, such as light signal transduction, hormone synthesis, glandular and root hair development, and stress [46–49].

In this study, the WGCNA method was used for the first time in *Dendrobium*, and the modules related to specific tissues and genes related to specific traits were identified. Hub genes were further analyzed to find related genes and predict gene functions [30]. Combining the WGCNA method and RNA-Seq data can be used to better mine the genes and transcription factors related to traits. In strawberry, modules related to tissue specificity such as torus were found in strawberry, and 7 hub genes were identified in torus tissue [37]; in tomato, genes related to vitamin C biosynthesis were found [50]; co-expression modules related to acidity and genes related to anthocyanin synthesis were found in apple [33,51]. In addition, specific modules in other tissues such as roots, leaves, flowers, and fruits at other periods can be excavated to find relevant metabolic processes and important genes and potential transcription factors.
