1 Introduction

When it comes to the "core role" influencing wheat yield, the development of panicles undoubtedly ranks among the top. After all, the appearance traits such as the length of the ear, the number of small flowers, and the density of grains per ear are directly related to how much grain can be harvested in the end. The differences in the developmental stages of the panicles not only affect whether they look good or not, but also determine the potential for the yield of individual wheat plants. Especially in the current situation where global attention to food security is increasing, even a slight increase in grain count per grain may bring considerable potential for increased production (Wang et al., 2017; VanGessel et al., 2022; Wei et al., 2022).

However, relying solely on phenotypic observation is clearly insufficient. The genetic background of wheat is complex, and the hexaploid genome makes it difficult for traditional methods to clearly understand the regulatory details. At this point, co-expression network analysis becomes particularly crucial. It does not directly target a single gene to "fight alone", but rather explores groups of genes that "appear" together in the same developmental sequence. Through the integration of transcriptome data, such methods can sort out several "meaningful" modules from a large amount of expression information, and then identify key pathways and hub genes. Many studies have successfully established the connection between spike development and yield traits using this approach, providing a new perspective for breeding (Li et al., 2022).

This study employed weighted gene co-expression network analysis (WGCNA) to identify gene modules related to wheat spike development, aiming to reveal candidate genes and regulatory networks that affect spike formation and yield. By integrating the transcriptome data during the development of wheat ears, this study aims to clarify the molecular mechanism of ear morphogenesis and provide genetic targets for breeding programs. The research results are expected to enhance the understanding of wheat inflorescence biology and increase wheat yield through molecular breeding strategies.

2 Molecular Regulatory Basis of Wheat Spike Development

2.1 Main developmental stages and regulatory genes of spike development

How do ears grow step by step? In fact, the process is not simple. From the initial double-ridge stage to the formation of small flower primordia and finally to the differentiation of spikelets, different things are happening at each stage, involving different genes "coming on stage". Genes like TaSPL15, WFZP, VRN1 and TaHOX4 have been repeatedly demonstrated by studies to play a crucial role in the development of floral organs and the formation of meristem (Figure 1). Once these genes mutate, the number of spikelets or the length of the spikelets may be problematic. Although these structural changes seem to be only "shape" changes, they actually directly affect the output behind them (Li et al., 2020; Ai et al., 2024).

Figure 1 Landscape of gene transcription, OCRs, and histone modifications in young spikes at the DRS and FPS (Adopted from Ai et al., 2024)

2.2 Key pathways involved in floral organ formation and spike differentiation

In the "behind-the-scenes system" that regulates the formation of wheat ears, not only individual genes are at play, but also many signaling pathways are working in coordination. For instance, the gibberellin pathway is not only related to the length of the spike but also affects the plant height. And behind this, there are also Q genes and others playing a role in regulation. However, this mechanism is far more than just hormones; a large number of transcription factors are also involved, such as TaMYB30, TaWRKY37, TaMYC2, etc. They not only regulate expression but are also "connected in series" with epigenetic mechanisms, acting like switches to control the activity level of genes in the spike. External environmental changes will also be taken into consideration, thereby jointly affecting the differentiation and development of floral organs (Lin et al., 2024; Liu et al., 2025; Yang et al., 2025).

2.3 Stage-specific gene expression and functional annotation

Not all genes are equally "busy" at all stages. Transcriptome data show that some genes are only highly expressed at specific stages, especially those related to hormones, metabolism and stress responses. The "main force" regulating these processes - transcription factor families such as bHLH, bZIP, MADS-box, MYB, NAC, and WRKY - are frequently mentioned in many studies (Zhu and Wang, 2025). They are responsible for maintaining the state of meristem, controlling the flowering time, and guiding the development of floral organs. It should be noted that, in addition to protein-coding genes, some long non-coding Rnas are also involved in the morphological regulation of the spike, and the regulatory process is much more complex than we imagined (Cao et al., 2021; Tan et al., 2025; Xu et al., 2025).

3 Methods and Strategies for Constructing Gene Co-expression Networks

3.1 Acquisition and preprocessing of transcriptome data (e.g., RNA-seq)

To figure out whether genes are "working together", the first step usually cannot do without transcriptome data. RNA-seq is now the most widely used, and it can simultaneously measure the expression levels of tens of thousands of genes. But once the data is in hand, it cannot be used directly. Quality checks must be conducted first to remove those "unclear" passages. Normalization cannot be omitted either; otherwise, the sequencing depth differences among different samples will be too great to compare. There are still some low-expression genes that make little contribution and need to be screened out first. Although these preprocesses may sound cumbersome, if they are not done or not done well, the network built later will be unreliable (Dam et al., 2017; Hou et al., 2022).

3.2 Principles of constructing weighted gene co-expression network analysis (WGCNA)

The WGCNA method, in the final analysis, is about looking at the "synchronization rate" of gene expression. The commonly used indicator is the Pearson correlation coefficient, but now some people also use more flexible ones, such as distance or Hellinger coefficient. These values are first transformed into a matrix of "how familiar who is with whom", and then a thing called topological overlap is calculated, which can reflect the degree of overlap in the "circle of friends" between one gene and other genes. Ultimately, we can draw out some closely connected small groups, which are called modules. These modules are often associated with certain biological functions and are the focus of the analysis (Zhang and Wong, 2022).

3.3 Module identification, sample clustering, and module specificity analysis

The step of dividing the modules is actually quite similar to seeing who is most familiar with whom on a social network. By using the algorithm of hierarchical clustering combined with dynamic tree cutting, hundreds or even thousands of genes can be separated into "circles of friends". Sometimes we also cluster the samples to see which ones have similar expression patterns and can correspond to the same developmental stage or a certain phenotype. The meaning of module-specific analysis is to match these "friend circles" with actual traits, such as whether they are particularly active during panicle development? In this way, we can step by step identify the regulatory modules most relevant to the target trait and also discover potential key genes, laying the foundation for the subsequent functional verification (Ruan et al., 2010; Gysi et al., 2020; Morabito et al., 2023).

4 Functional Annotation and Module–Trait Association Analysis

4.1 Correlation between modules and agronomic traits such as spike length, branch number, and floret number

When looking at the relationship between modules and agronomic traits, there is a problem that is often overlooked: the connection between expression patterns and traits is not always immediately apparent. However, once you substitute the yield-related traits such as panicle length, branching number, and flower number into the analysis, it will be found that some co-expression modules are indeed closely linked to these indicators (Wei et al., 2022; Yang et al., 2023). Which "promising" genes are there in these modules exactly? Through module-trait pairing, researchers can identify the cluster that may be closely related to the spike structure as the focus for subsequent functional verification.

4.2 GO annotation and KEGG pathway enrichment analysis

After identifying the relevant modules, it is not immediately known what they are doing. At this point, one has to rely on annotation tools like GO and KEGG. The results were not too unexpected either: Many module genes were concentrated in the categories of hormone signal transduction, developmental regulation, and transcription factor activity, which were closely related to panicle development (Li et al., 2019; 2022). These enriched information are equivalent to "labeling" the module. Although they do not represent all the functions, they can at least provide a starting point for interpretation.

4.3 Identification of key hub genes within modules and construction of regulatory networks

The genes in the module are not equal. Some genes have particularly large "social circles" and are related to many other genes, and are thus called hub genes. Screening out these hubs is to then draw the network diagram to see if they are the core nodes on the regulatory chain (Wang et al., 2017; VanGessel et al., 2022). This kind of network analysis not only helps us understand how genes check and balance each other or cooperate, but also makes it convenient to pick out candidate targets worth using for transgenic or editing experiments. For those engaged in variety improvement, these genes might be the most valuable breakthrough points.

5 Analysis of Transcription Factors and Regulatory Networks

5.1 Identification of transcription factor families associated with spike development modules (e.g., MADS-box, bHLH)

In fact, in the panicle development module, what was first noticed was not a specific gene, but some clearly concentrated families of transcription factors - familiar faces like MADS-box and bHLH, which account for a large proportion in many modules. They are not the first time to be associated with the development of floral organs and meristem, but this co-expression network analysis has still made their roles somewhat clearer. Especially the MADS-box genes, during the gradual development of the spike, their activation sequence is very regular (VanGessel et al., 2022; Lin et al., 2024). Although such families have long been studied, their continuous screening out in the spike-related modules indicates that their main regulatory role still cannot be ignored.

5.2 Expression pattern analysis between TFs and their target genes within modules

It's not enough to just look at the transcription factors; you also need to see who they partner with. By analyzing the expression patterns of transcription factors and their potential target genes, researchers found that many of them were simultaneously upregulated at the same developmental time point. This rhythmic coordination suggests that they may have a "partner" relationship in spike initiation and differentiation (Xu et al., 2025). Interestingly, these pairings are not scattered but concentrated in specific modules, indicating that the regulatory programs within these modules are very tight. Compared with the spatio-temporal expression data, this phenomenon of expression synchronization is also more convincing, helping us gradually piece together the regulatory map of panicle development.

5.3 Signal transduction pathways and regulatory mechanisms mediated by TFs

But the story of transcription factors is far more than just "up-regulation and down-regulation". In many cases, they are also caught between hormone signals and gene expression, playing the role of "translators". Familiar plant hormones such as auxin and cytokinin often exert their regulatory effects through TF (Li et al., 2019). Of course, the pathways mediated by different TFS are not exactly the same. Some control the morphological construction of the spike, while others may be more inclined towards organ-level transformation. All this information, when put together, helps explain how the final structure of the spike was gradually "determined" and also points out possible genetic targets for subsequent yield improvements.

6 Case Study: Empirical Validation of Key Modules and Genes

6.1 Correlation analysis of the “turquoise” module with spike length regulation

In the research, panicle length is regarded as one of the main factors affecting wheat yield. However, not all expression modules are related to this trait. The peculiarity of the "turquoise" module lies in that multiple genes within it exhibit expression patterns significantly related to panicle length in different wheat varieties (Wei et al., 2022; Yang et al., 2023). From the perspective of co-expression analysis, this module may be involved in the elongation process and structural regulation of the spike (Figure 2). Although the current evidence is mainly at the level of correlation, its research value is already very clear - it is worth further exploration.

Figure 2 Co-expression module associated with spike length (SL) (Adopted from Yang et al., 2023)

6.2 qRT-PCR validation and spatiotemporal expression of candidate hub genes (e.g., TaFUL2)

Network analysis alone is not enough; we need to take a look at the actual expression situation. Some hub genes in the "turquoise" module represented by TaFUL2 were verified for expression by qRT-PCR. The results showed that these genes were active at the key stages of spike development, especially in spike tissues (Lin et al., 2024; VanGessel et al., 2022). This spatio-temporal expression characteristic not only confirms the speculation of network analysis but also further indicates that these genes are indeed involved in the regulation of inflorescence structure. Although it is only a preliminary verification, the direction is already very clear.

6.3 Preliminary functional evaluation through gene editing or transformation experiments

At the functional verification stage, the researchers did not remain at the expression level but directly conducted knockout or overexpression experiments on key genes such as TaFUL2 using CRISPR. The results are clear at a glance. Both panicle length and the number of spikelets have been affected (Wang et al., 2017; Lin et al., 2024). Although these experiments are still in their early stages, they have already provided strong functional hints. Ultimately, these verifications give us more confidence to apply the predictions in the co-expression network to practical genes that can truly be used in wheat breeding.

7 Conclusion and Perspectives

When studying the development mechanism of wheat ears, in the final analysis, it is still impossible to avoid a large number of gene regulatory factors that affect the shape and yield of the ears. In this study, through network analysis, we identified some co-expression modules and also found some key gene clusters closely related to spikelet formation and spike axis elongation. These things have not been mentioned for the first time, but now their dynamic changes during the inflorescence development process are a bit clearer. Although it cannot be said to have been fully mastered yet, at least the framework has been established, which has opened up new ideas for further in-depth understanding of the regulation of spike structure and can also be regarded as providing some molecular basis for yield improvement.

But the problem is not small either. The genome of wheat is large enough and polyploid, with information piled up layer upon layer. The integration of omics data is by no means an easy task. The transcriptome data from different materials and at different time points vary significantly. Coupled with the fact that most of the candidate genes have not yet had time for functional verification, it is still hard to say whether many co-expression modules are truly useful at present. Not to mention the data at the epigenetic regulation, protein and metabolic levels, the current binding with the transcriptome is still far from sufficient, which leaves a piece of the puzzle missing in the regulatory map of the entire spike development.

So, the next focus might have to be placed on cross-omics integration. Relying solely on a single data dimension is no longer far. By aggregating the information from the transcriptome, epigenome and genome, and combining it with a more mature network analysis model, it might be possible to build a more realistic regulatory map of spike development. Moreover, if these networks can be combined with CRISPR editing and functional verification experiments to advance simultaneously, there might be a real breakthrough in breeding. After all, the ultimate goal is still to transform these molecular-level understandings into the cultivation of wheat varieties with high yields and excellent ear shapes. The key lies in their implementation.

Acknowledgments

We thank Mr Z. Wu from the Institute of Life Science of Jiyang College of Zhejiang A&F University for his reading and revising suggestion.

Conflict of Interest Disclosure

The authors affirm that this research was conducted without any commercial or financial relationships that could be construed as a potential conflict of interest.

References

Ai G., He C., Bi S., Zhou Z., Liu A., Hu X., Liu Y., Jin L., Zhou J., Zhang H., Du D., Chen H., Gong X., Saeed S., Su H., Lan C., Chen W., Li Q., Mao H., Li L., Liu H., Chen D., Kaufmann K., Alazab K., and Yan W., 2024, Dissecting the molecular basis of spike traits by integrating gene regulatory networks and genetic variation in wheat, Plant Communications, 5(5): 100879.

https://doi.org/10.1016/j.xplc.2024.100879

Cao P., Fan W., Li P., and Hu Y., 2021, Genome-wide profiling of long noncoding RNAs involved in wheat spike development, BMC Genomics, 22: 493.

https://doi.org/10.1186/s12864-021-07851-4

Dam S., Võsa U., Van Der Graaf A., Franke L., and Magalhães J., 2017, Gene co-expression analysis for functional classification and gene–disease predictions, Briefings in Bioinformatics, 19: 575-592.

https://doi.org/10.1093/bib/bbw139

Gysi D., De Miranda Fragoso T., Zebardast F., Bertoli W., Busskamp V., Almaas E., and Nowick K., 2020, Whole transcriptomic network analysis using co-expression differential network analysis (CoDiNA), PLoS ONE, 15(10): e0240523.

https://doi.org/10.1371/journal.pone.0240523

Hou J., Ye X., Feng W., Zhang Q., Han Y., Liu Y., Li Y., and Wei Y., 2022, Distance correlation application to gene co-expression network analysis, BMC Bioinformatics, 23: 46.

https://doi.org/10.1186/s12859-022-04609-x

Li J., Jiao Z., He R., Sun Y., Xu Q., Zhang J., Jiang Y., Li Q., and Niu J., 2019, Gene expression profiles and microRNA regulation networks in tiller primordia stem tips and young spikes of wheat Guomai 301, Genes, 10(9): 686.

https://doi.org/10.3390/genes10090686

Li Y., Li L., Zhao M., Guo L., Guo X., Zhao D., Batool A., Dong B., Xu H., Cui S., Zhang A., Fu X., Li J., Jing R., and Liu X., 2020, Wheat frizzy panicle activates Vernalization1‐A and Homeobox4‐A to regulate spike development in wheat, Plant Biotechnology Journal, 19: 1141-1154.

https://doi.org/10.1111/pbi.13535

Li Z., Hu Y., Ma X., Da L., She J., Liu Y., Yi X., Cao Y., Xu W., Jiao Y., and Su Z., 2022, WheatCENet: a database for comparative co-expression networks analysis of allohexaploid wheat and its progenitors, Genomics Proteomics and Bioinformatics, 21: 324-336.

https://doi.org/10.1016/j.gpb.2022.04.007

Lin X., Xu Y., Wang D., Yang Y., Zhang X., Bie X., Gui L., Chen Z., Ding Y., Mao L., Zhang X., Lu F., Zhang X., Uauy C., Fu X., and Xiao J., 2024, Systemic identification of wheat spike development regulators by integrated multi-omics transcriptional network GWAS and genetic analyses, Molecular Plant, 17(3): 438-459.

https://doi.org/10.1016/j.molp.2024.01.010

Liu P., Xue S., Jia J., Zhao G., Liu J., Hu Y., Kong C., Yan D., Wang H., Liu X., Lu Z., and Gao L., 2025, The wheat transcription factor Q functions in gibberellin biosynthesis and signaling and regulates height and spike length, The Plant Cell, 37(8): koaf183.

https://doi.org/10.1093/plcell/koaf183

Morabito S., Reese F., Rahimzadeh N., Miyoshi E., and Swarup V., 2023, hdWGCNA identifies co-expression networks in high-dimensional transcriptomics data, Cell Reports Methods, 3(6): 100498.

https://doi.org/10.1016/j.crmeth.2023.100498

Ruan J., Dean A., and Zhang W., 2010, A general co-expression network-based approach to gene expression analysis: comparison and applications, BMC Systems Biology, 4: 8.

https://doi.org/10.1186/1752-0509-4-8

Tan B., Xie Y., Peng H., Wang M., Zhu W., Xu L., Cheng Y., Wang Y., Zeng J., Fan X., Sha L., Zhang H., Qin P., Zhou Y., Wu D., Li Y., and Kang H., 2025, Transcriptome profiling of spike development reveals key genes and pathways associated with early heading in wheat–Psathyrstachys huashanica 7Ns chromosome addition line, Plants, 14(13): 2077.

https://doi.org/10.3390/plants14132077

VanGessel C., Hamilton J., Tabbita F., Dubcovsky J., and Pearce S., 2022, Transcriptional signatures of wheat inflorescence development, Scientific Reports, 12: 17224.

https://doi.org/10.1038/s41598-022-21571-z

Wang Y., Yu H., Tian C., Sajjad M., Gao C., Tong Y., Wang X., and Jiao Y., 2017, Transcriptome association identifies regulators of wheat spike architecture1, Plant Physiology, 175: 746-757.

https://doi.org/10.1104/pp.17.00694

Wei J., Fang Y., Jiang H., Wu X., Zuo J., Xia X., Li J., Stich B., Cao H., and Liu Y., 2022, Combining QTL mapping and gene co-expression network analysis for prediction of candidate genes and molecular network related to yield in wheat, BMC Plant Biology, 22: 127.

https://doi.org/10.1186/s12870-022-03677-8

Xu X., Lin H., Zhang J., Burguener G., Paraiso F., Tumelty C., Li C., Liu Y., and Dubcovsky J., 2025, Spatial and single-cell expression analyses reveal complex expression domains in early wheat spike development, bioRxiv, 26(1): 352.

https://doi.org/10.1101/2025.02.15.638402

Yang G., Pan Y., Cui L., Chen M., Zeng Q., Pan W., Liang Z., Edwards D., Batley J., Han D., Deng P., Yu H., Henry R., Weining S., and Nie X., 2023, Genetic basis of expression and splicing underlying spike architecture in wheat (Triticum aestivum L.), bioRxiv, 2023: 05.

https://doi.org/10.1101/2023.05.04.539218

Yang Z., Bai W., Guo G., Huang S., Wang Y., Zhou Y., Zhang Y., and Sun J., 2025, The Q-interacted protein QIP3 recruits TaTPL to regulate spike architecture in wheat., The Plant Journal, 122(2): e70149.

https://doi.org/10.1111/tpj.70149

Zhang T., and Wong G., 2022, Gene expression data analysis using Hellinger correlation in weighted gene co-expression networks (WGCNA), Computational and Structural Biotechnology Journal, 20: 3851-3863.

https://doi.org/10.1016/j.csbj.2022.07.018

Zhu S.J., and Wang W., 2025, Functional validation of maize phosphate transporters using overexpression lines, Maize Genomics and Genetics, 16(4): 219-228.

https://doi.org/10.5376/mgg.2025.16.0019