Optimizing reference gene selection for accurate gene expression analysis in plants Authors: Zili Xiong, Cunnian Fu, Shaoyong Huang, Jianlei Shi Published: 2025 ABSTRACT ABSTRACTAccurate selection of reference genes is crucial for reliable gene expression analysis in plants. Traditional reference genes, such as GAPDH and ACT, are widely used but often show variable stability under different conditions, stresses, tissue types, and developmental stages. Recent advances in multi-omics technologies, including transcriptomics, offer new approaches for improving reference gene selection. These methods allow for the integration of diverse datasets to identify genes with stable expression across various environmental stresses and developmental stages, providing more robust and context-specific normalization controls. High-throughput sequencing technologies, such as RNA-seq, have enabled the large-scale identification of stable reference genes, and further enhanced this process by correlating gene expression at the transcript level. Additionally, the application of computational tools helps researchers optimize reference gene selection, making the process more efficient and standardized. Personalized and condition-specific reference gene databases are emerging as valuable resources for selecting the most appropriate genes based on experimental conditions. This paper explores the current trends and challenges in reference to reference gene selection and the potential of a transcriptomics approach to address these challenges. The use of these advanced methods will increase the accuracy and reliability of plant gene expression studies, accelerate discoveries in crop improvement and stress resilience in plant biology and agricultural sciences.KEYWORDS: Reference genes, expression analysis, multi-omics, transcriptomics, qRT-PCR Introduction Accurate gene expression analysis is fundamental for understanding plant development, metabolic, and stress responses in plant molecular biology research.1 Real-time quantitative PCR (qRT-PCR) is one of the most reliable and widely applied techniques, known for its high sensitivity and quantitative precision.2 However, the accuracy of qRT-PCR results were influenced by RNA integrity, reverse transcription efficiency, and environmental variations.3 Therefore, normalization using validated reference genes is essential for reproducible and meaningful data interpretation.4 Reference genes are endogenous genes that exhibit stable expression across different tissues, developmental stages, and environmental conditions.5 These genes are involved in basic cellular metabolism and structural maintenance processes and, in theory, are not affected by experimental treatments or environmental factors. Classic reference genes such as GAPDH, ACT, and 18S rRNA have been widely adopted for normalization; however, increasing evidence indicates that their stability is context-dependent.5 With the rapid expansion of transcriptomic and multi-omics datasets, large-scale screening of condition-specific reference genes has become feasible, improving both accuracy and reproducibility. Under different stresses or tissues, expression of these classic reference genes fluctuates significantly.6–9 Particularly under environmental stress conditions (such as drought, salinity, low temperature, and pests), the expression stability of different reference genes shows notable differences, which is particularly significant in studies of stress/development responsive gene expression.10-15 For example, under salt stress, the expression stability of ACT and 18S rRNA is poor in certain plants, while some novel candidate reference genes, such as eIF4α and EF1α, show higher stability across multiple stress conditions.10-16 Therefore, systematically screening for suitable reference genes under specific conditions is crucial for ensuring the reliability of results. At present, various software tools and methods (such as GeNorm, NormFinder, and BestKeeper) have been developed for evaluating the stability of reference genes.2,17,18 These tools allow researchers to identify reference genes with stable expression under specific conditions based on experimental data, thereby reducing errors caused by unstable reference gene expression.2,17,18 Additionally, with the development of high throughput sequencing technologies, an increasing amount of omics data is being used to identify new candidate reference genes.5,14,19 For instance, using transcriptomic data to screen reference genes has made significant progress in plants such as rice and wheat, providing valuable resources for the selection of reference genes in different species.7,19 The present review reviews the recent research progress in the selection of plant reference genes, focusing on the expression stability of reference genes in different plant species, tissue types, developmental stages, and stress conditions. By analyzing the selection strategies and evaluation methods for different reference genes, we provide a reference for the screening of reference genes in plant molecular biology research and discuss the potential applications of reference gene selection in multi-omics data and molecular marker-assisted screening in the future.20,21 Definition and role of plant reference genes Reference genes are usually involved in essential biological processes such as glycolysis, cytoskeleton formation, and ribosome synthesis.5 In gene expression studies, reference genes are commonly used as normalization controls to correct for RNA extraction efficiency, reverse transcription efficiency, and errors in qRT-PCR, thereby improving the reliability and accuracy of the data.22-25 An ideal reference gene should maintain a stable expression level under different conditions and should not change even under environmental stress or treatment.26-28 However, recent evidence suggests that even traditional housekeeping genes such as ACT and GAPDH exhibit instability under specific abiotic or biotic stress conditions.15,16 Therefore, the concept of a “universal reference gene” has been replaced by the need for context-specific validation. To minimize bias, modern studies often combine multiple reference genes and evaluate their stability statistically using algorithms such as GeNorm, NormFinder, and BestKeeper.2,17,18 Commonly used plant reference genes and their functions Commonly used reference genes include structural protein-coding genes such as ACT and α-tubulin (TUA), glycolysis-related genes such as GAPDH, translation-related genes such as EF1α, and ribosomal RNA genes (e.g., 18S and 26S rRNA).10,29,30 These genes play key roles in cellular metabolism and are considered to have stable expression under various conditions (Table 1). ACT and TUA are structural proteins widely used in plant gene expression research. ACT plays a key role in cytoskeleton formation and cell movement, while TUA is a component of microtubules and is essential for cell division and elongation. Owing to their important roles in maintaining and dynamically changing cell structure, these genes are used as stable references in various plants (e.g., Arabidopsis, rice, tomato, wheat, cucumber, citrus, cotton).10-12,29,30 However, some studies have shown that under certain environmental stresses (e.g., salt stress, drought), the expression of these structural genes is not always stable. For example, Czechowski et al.5found that ACT expression fluctuated under drought stress in Arabidopsis, suggesting that ACT is not always an ideal reference gene under such conditions. Therefore, when ACT and TUA are selected as reference genes under specific stress or conditions tested, further validation of their expression stability is required. Table 1.Overview of reference genes and their relative stability under different stresses.Reference geneMain functionRepresentative speciesStability under stress conditions (Summary)Recommended use contextKey observations/controversiesReferences ACTIN Cytoskeleton, cell divisionArabidopsis, tomato, riceOften unstable under biotic stress (e.g., pathogen infection); variable under drought; moderate stability under abiotic stressNot recommended for pathogen or wound studiesFrequently overused despite known instability; some studies report acceptable performance in non-stressed tissues[5,31,32] GAPDH Glycolysis, energy metabolismArabidopsis, tomatoDown-regulated under salt, drought, osmotic stress; highly variable in stress conditionsAvoid in abiotic stress experimentsDespite common use, multiple studies show poor stability; may be stable in control conditions only[5,12] EF1α Protein synthesis (translation elongation)Grapevine, cucumberGenerally stable under drought, salinity; slight variation under heat stressSuitable for drought/salinity studiesHigh consistency across studies; one of the most reliable among classical candidates[10,33] UBQ10 Protein ubiquitination, protein turnoverTomato, RiceBroadly stable across diverse stresses; minor fluctuations under prolonged heatPreferred for multi-stress or long-term experimentsConsistently ranked among top stable genes in qPCR validation studies[34,35] 18S rRNA Ribosome assembly translationPeach, zucchiniStable in many conditions, but shows variability under nutrient deficiency and extreme temperaturesUse with caution in nutrient-limited or low-light environmentsOften overestimated as universal reference; not suitable for gene expression normalization in metabolic studies[36,37] Overview of reference genes and their relative stability under different stresses. GAPDH and EF1α are also commonly used reference genes, particularly in rapidly metabolizing plant cells.10,15 GAPDH is involved in glycolysis and plays a crucial role in cellular energy metabolism, so it is considered to have relatively stable expression under various tissues and treatment conditions. EF1α acts as an elongation factor in protein translation, aiding the transfer of amino acids, and is a necessary component of protein synthesis. Li et al.15 found that GAPDH and EF1α exhibited relatively stable expression levels under different stress conditions in celery. In addition, ribosomal RNA genes (e.g., 18S and 26S rRNA) are also commonly used as references owing to their high abundance in cells and are usually considered to express stably. However, recent studies have shown that ribosomal RNA expression fluctuates under stress conditions in some plants, indicating that ribosomal RNA genes also need to be validated when used as reference genes.38 Challenges in selecting plant reference genes The stability of plant reference genes is crucial for accurate gene expression analysis. However, differences in different conditions, plant tissues, and species significantly affect the stability of reference genes. Therefore, selecting appropriate reference genes under different conditions remains a major challenge in plant molecular biology research. We analyzed the impact of stress conditions, tissue and developmental stage differences, and interspecies differences. Impact of stress conditions When plants are subjected to different types of stress (e.g., salt stress, drought, low temperature, and pathogen infection), complex physiological and biochemical responses are triggered, leading to significant changes in gene expression patterns.5,12 Therefore, the expression of common reference genes was unstable under these stress conditions. For example, GAPDH, a metabolic enzyme gene, shows relatively stable expression under normal growth conditions, but studies have shown that its expression is significantly affected under salt and drought stress.6,16 Additionally, traditional reference genes such as ACT and 18S rRNA also exhibit expression fluctuations when plants face abiotic stresses.29,33 These examples highlight the need to optimize the choice of reference genes under specific stress conditions. Under drought, salinity, and pathogen stress, the stability of traditional reference genes such as ACT, GAPDH, and 18S rRNA often decreases.6,16 In contrast, EF1α and eIF4α has shown superior stability across multiple stresses in celery and cucumber.10,15 Therefore, condition-specific validation has become the most reliable strategy. Differences in tissues and developmental stages There are significant differences in gene expression patterns across different tissues and developmental stages in plants. The expression levels of the same gene vary between tissues or developmental stages, which limits the applicability of common reference genes. For example, in a study on rice development, genes such as ACT and GAPDH exhibited differences in expression at different developmental stages, making them less suitable as reference genes in rice tissues at different stages of development.6 Similarly, in Arabidopsis, some studies have shown that 18S rRNA expression varies between leaves and roots, potentially affecting its stability as a reference gene in different tissues.5 To address the impact of tissue and developmental stage on reference gene selection, researchers typically need to screen and validate reference genes suitable for specific tissues or developmental stages before the experiment. For instance, Tong et al.37found that TEF2 maintained stable expression during fruit development in peach, making it an ideal reference gene. Additionally, Saha et al.39 systematically screened and identified a set of specific reference gene combinations in millet leaves, stems, and roots to ensure accuracy in tissue-specific expression analysis. In conclusion, owing to variations in gene expression across tissues and developmental stages, appropriate reference gene selection should consider the specific tissues or developmental stages involved in the experiment to improve the reliability of expression analysis. Interspecies differences The diversity of plant species means that a reference gene that is stable in one species exhibits unstable expression in another.29,30,33 Differences in gene expression regulation, metabolic pathways, and physiological states across species limit the stability of reference genes across species. For instance, ACT and TUB are widely used in Arabidopsis and rice, but in other species (e.g., tomato), their expression stability differs.11,12 Additionally, with the increasing number of non-model plant genomes being sequenced, researchers have found that some reference genes that are stable in model plants show significant fluctuations in new species.35,36 For example, in potato research, traditional reference genes such as ACT and GAPDH show expression differences under certain environmental conditions, suggesting that they were not ideal reference genes in all species.40 Recent advances in reference gene selection With the development of genomic sequencing technologies and molecular biology tools such as quantitative PCR, progress has been made in the selection of reference genes. The expression stability of reference genes under different conditions and plant types is crucial for ensuring the accuracy of gene expression analysis. The stability of reference genes varies depending on the plant type and specific conditions, making it necessary to select appropriate reference genes tailored to specific plants and conditions. Reference gene selection in model plants Model plants, such as Arabidopsis and rice, are widely used in plant molecular biology research, and the selection of reference genes in these plants has been studied in depth.41 Arabidopsis is an important model plant in molecular plant research, and studies have shown variations in the expression stability of its traditional reference genes under different conditions.5 Czechowski et al.5 performed a high-throughput transcriptomic analysis and identified a set of reference genes suitable for various conditions, such as UBQ10 and EF1α. This data-driven large-scale screening method has improved the reliability of reference gene selection. Additionally, in rice, Jain et al.6 conducted a comparative analysis of the stability of multiple reference genes and found that EF1α was relatively stable under salt and drought stress, making it suitable for use as a reference gene in qRT‒PCR. In recent years, RNA-seq has provided quantitative expression data for thousands of genes simultaneously. By computing expression variance (CV values) across datasets, researchers have identified stably expressed genes.5,14 These approaches outperform traditional candidate-gene screening, particularly under complex stress combinations. Reference gene selection in crops Owing to the diversity of growing environments and cultivation conditions, the selection of reference genes becomes an important challenge for gene expression analysis in staple crops. However, significant progress has been made in the selection of reference genes for crops such as wheat, maize, and soybean (Table 2). For example, in rice, Moraes et al.42 analyzed the stability of reference genes in different tissues and stress conditions and found that EF1α and UBQ5 exhibited good stability under various environmental conditions, making them suitable for gene expression analysis. Similarly, in studies on drought stress in wheat, researchers have reported that CJ705892, ACT, and UBI exhibit high expression stability, making them suitable reference genes.43 Table 2.List of stable reference genes identified in major crops.Plant speciesStable reference genesExperimental conditions testedKey functional roles & stability driversPerformance summary & recommendationsReferencesRice EF1α, UBQ5, eIF4A Germination, salt stress, heat stress, developmental stagesInvolved in translation initiation and protein synthesis; generally conserved expression across tissues and stressesHighly recommended for abiotic stress studies; EF1α and UBQ5 show consistent stability across multiple conditions[42,44]Maize EF1α, EIF4A Abiotic stress, hormone treatments, pathogen infectionCore components of translational machinery; low regulatory plasticity under stressReliable across diverse treatments; suitable for multi-stress or hormonal studies[45,46]Wheat Ta2291, TaFNRII, CJ705892 Drought and nutrient treatmentsUnknown functions (some are novel); may be regulated by stress-responsive pathwaysNovel genes with broad stability; promising candidates for future validation in related cereals[7,43,47]Soybean TIP41, F-box Salt, drought, heat, ABA treatmentsTIP41: involved in vesicle trafficking; F-box: ubiquitin ligase componentBoth exhibit minimal expression variation under stress; potential for use in legume stress research[23,48]Cotton UBQ14, PP2A1, UBQ7 Tissues and developmental stagesUbiquitination-related (UBQ), phosphatase activity (PP2A)-essential housekeeping functionsStrongly stable across developmental and environmental gradients; ideal for gene[49,50] List of stable reference genes identified in major crops. In soybean research, owing to the complex biotic and abiotic stresses it faces, the selection of reference genes has been more extensively studied. Hu et al.48 compared the expression stability of multiple reference genes under various stress conditions and found that TIP41, UKN1 and UKN2 exhibited stable expression when the light quality was changed. Additionally, new analytical tools, such as Ref-Finder and GeNorm, have been applied to the selection and stability analysis of crop reference genes. These tools use multifactor analysis of gene expression data to help researchers more accurately select appropriate reference genes, thereby improving the reliability of experiments.2,51 Reference gene selection in horticultural crops Horticultural crops such as tomatoes, grapes, and strawberries exhibit gene expression differences across different tissues and developmental stages. Therefore, selecting suitable reference genes is particularly important in the study of horticultural crop species. For example, during fruit development in tomatoes, gene expression differences across different developmental stages are significant. ExpósitoRodríguez et al.11 compared the expression of multiple reference genes at different fruit development stages and found that TIP41 and SAND exhibit high expression stability during fruit development, making them ideal reference genes. In grapevine, owing to the differences in stress responses among different cultivars, reference gene selection has become a major issue. Studies have shown that GAPDH and EF1α exhibit good expression stability across different tissues and stress conditions in grapes, making them widely used for gene expression normalization in grapes.13 Furthermore, in the tissue-specific expression analysis of strawberries, GAPDH and ACTIN1 exhibit consistent expression levels, providing reliable references for strawberry gene function studies.52 These studies indicate that reference gene selection for different horticultural crops needs to consider the specific tissues and developmental stages involved to ensure the accuracy of gene expression analysis. Selection of new reference gene in plant species With the development of high-throughput sequencing and multi-omics technologies, reference gene selection methods have continuously innovated, offering greater accuracy for plant gene expression analysis. Gene chip/EST-based screening and the application of multi-omics methods provide new perspectives on reference gene selection, making the process more scientific and efficient (Figure 1). Traditional reference gene selection typically relies on the validation of a few candidate genes, but these genes often exhibit significant variability in expression stability under different conditions and in different plant species. In recent years, RNA-seq has revolutionized reference gene screening. By analyzing transcriptomic data, researchers have identified reference genes that exhibit stable expression across various conditions and tissues. For example, through transcriptomic data analysis, Czechowski et al.5 identified a set of reference genes in Arabidopsis that exhibited stable expression under different stress conditions. This large-scale screening approach evaluates the stability of candidate reference genes by calculating the coefficient of variation across different samples, thus avoiding the instability caused by single reference genes influenced by specific conditions. Identification of suitable reference genes for qRT-PCR normalization. Biological samples from different tissues, developmental stages, or experimental treatments are collected. RNA is extracted and analyzed by either a DNA microarray or RNA-Seq. Candidate reference genes were selected, and specific primers were designed for qRT-PCR. Amplification plots and melt-curve analyses were performed to verify primer specificity and amplification efficiency. The stability of the candidate reference genes was evaluated using GeNorm, NormFinder, and BestKeeper. Based on these analyses, the most stable reference genes are identified for accurate normalization of target gene expression. In important crops such as pepper, transcriptomic technologies have also been widely applied to reference gene screening. For example, Cheng et al.53 screened a set of stable reference genes, such as CaREV05 and CaREV08, in chili peppers under different growth stages and stress conditions using RNA-seq data. This data-driven screening method effectively covers different growth environments and different treatments, improving the generalizability and reliability of reference genes. Moreover, transcriptomic analysis also reveals differences in expression stability between tissues and developmental stages, making the selection of tissue-specific or stage-specific reference genes more precise. Future perspectives In recent years, the methodological landscape for reference gene (RG) selection has shifted from small-scale empirical testing toward multi-omics-driven discovery. Classical tools such as geNorm, NormFinder, and BestKeeper evaluate expression stability within a limited gene set, but they are constrained by small sample size and experimental context. In contrast, transcriptome-wide and multi-omics analyses can now systematically identify stable candidate RGs by integrating gene expression variability across tissues, developmental stages, and stress conditions. For instance, transcriptome meta-analysis frameworks and databases such as Genevestigator’s RefGenes enable users to retrieve condition-specific RGs across thousands of RNA-seq or microarray datasets.19 The recently developed RGeasy web platform further operationalizes this approach by integrating transcriptomic data with statistical algorithms to automatically select RGs across multiple conditions.54 Moreover, integrative multi-omics studies that combine transcriptome and proteome layers have demonstrated that consistent expression profiles across data types improve RG reliability. These advances illustrate how data integration enhances reproducibility and reduces subjective bias in RG selection.55 Looking forward, artificial intelligence (AI) and machine learning (ML) are expected to transform RG identification into a predictive modeling process. A feasible computational pipeline could involve: (i) constructing a multi-omics compendium encompassing transcriptome, proteome, and epigenome data for the target species; (ii) applying ML algorithms such as random forest, support vector machine, or Lasso regression to evaluate expression stability based on quantitative features (e.g., coefficient of variation, expression breadth, network centrality, and tissue-specificity scores); and (iii) predicting context-specific RG panels optimized for distinct conditions (e.g., “salt-stressed leaves” or “flowering tissues”). Studies in human and plant transcriptomics have demonstrated that ML models outperform traditional stability metrics by integrating multidimensional data.41 Deep-learning-based multi-omics frameworks are also emerging to identify invariant expression signatures across stress treatments.56 In plant systems, combining these models with RNA-seq meta-analysis tools such as RefGenes and RGeasy could yield automated, interpretable, and condition-aware RG recommendations. We predict that future pipelines will integrate network-aware learning to penalize co-regulated genes, and transfer-learning architectures to extrapolate RG stability from model plants to less-studied species – ushering in a new era of data-driven, scalable, and reproducible reference gene selection. Acknowledgments This research was funded by grants from Zhejiang Provincial Major Agricultural Science and Technology Projects of New Varieties Breeding (2021C02065), Project of the Wenzhou Municipal Collaborative Group for Breeding New Agricultural Varieties (ZX20240022) and China Agriculture Research System of MOF and MARA (CARS23G45). Funding Statement This research was funded by grants from Zhejiang Provincial Major Agricultural Science and Technology Projects of New Varieties Breeding (2021C02065), Project of the Wenzhou Municipal Collaborative Group for Breeding New Agricultural Varieties (ZX20240022) and China Agriculture Research System of MOF and MARA (CARS23G45). Disclosure statement During the preparation of this manuscript, the authors used ChatGPT5 Thinking (OpenAI; accessed September 2025) to assist with the conceptualization and layout of Figure 1, as well as for English language editing to improve the clarity and readability of the text. The authors reviewed and edited the output and take full responsibility for the content of this publication. Author contributions Conceptualization: XZ and SJ; validation: XZ and FC; data curation: HS; writing original draft preparation: XZ and SJ; writing review and editing, SJ; funding acquisition, SJ; All authors have read and agreed to the published version of the manuscript. Conflict of interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. References