SRS: An intelligent and robust approach for confirmation of plant transcription factor–DNA interactions

Transcription factors (TFs), representing 5%–8% of eukaryotic nuclear genome, bind specific DNA sequences like promoters to regulate transcription (Lambert et al., 2018). Identifying these sequences is vital for understanding TF functions. Techniques such as chromatin immunoprecipitation sequencing (ChIP-Seq), electrophoretic mobility shift assay (EMSA), yeast one-hybrid (Y1H) assay, dual-luciferase reporter LUC/REN assay, and β-glucuronidase (GUS) reporter are used to validate TF–promoter interactions but require extensive instrumentation and chemicals (Abid et al., 2022; Park, 2009). An alternative, the RUBY/eYGFPuv assay, uses modified plant leaf colour as a visible, cost-effective method for studying DNA–protein interaction (Sun et al., 2023). Advances in genomics, including RNA sequencing and ChIP-Seq, underscore the need for efficient, reliable visual detection systems to map TF binding sites, crucial for elucidating their regulatory roles and broader biological impacts. To develop a visual reporter for TF–DNA interactions, we targeted genes influencing leaf colour by modulating chlorophyll (Chl) degradation. The Stay-Green1 (SGR1) gene, crucial for Chl breakdown during senescence, encodes magnesium dechelatase. Mutations in SGR1 result in a stay-green phenotype, while overexpression leads to yellowing (Shimoda et al., 2016). We chose SGR1 from 32 candidates, divided into three subgroups, and cloned SGR1 genes from Arabidopsis thaliana (AtSGR1), Oryza sativa (OsSGR1), and two from Ginkgo biloba (GbSGR1 and GbSGR1L) (Figure S1). Using the Cauliflower Mosaic Virus 35S promoter, we expressed these genes in Nicotiana benthamiana leaves via Agrobacterium tumefaciens-mediated transformation (Figure S2). All transformed areas exhibited accelerated yellowing, with AtSGR1 exhibiting the most rapid and significant Chl degradation, demonstrating its potential as a TF–DNA interaction reporter (Figure 1a–c). Additionally, enhancements including a Kozak consensus sequence for improved translation, darkness to stimulate yellowing, and maintaining temperatures between 22 and 25°C significantly boosted Chl degradation (Figure S3). To efficiently monitor Chl level changes, we developed the Smart Model for tracking Chl change from Green to Yellow (SMGY). This model utilizes a second-order polynomial regression and a colour difference correction matrix, calibrated against a standard colour chart to minimize image colour variations. We integrated a remote diagnosis system via the WeChat Mini Program for on-site, real-time, non-destructive Chl detection in plant leaves (Figure 1d). To predict SPAD values from images, we analysed 14 features with significant correlations (r > 0.5; Figure S4a; Table S1) and used a stacking ensemble of five machine learning models (Figure S4b; Table S2). After 100 iterations, the model achieved an R2 of 0.85, RMSE of 2.4, and NRMSE of 12.24% (Figure S4c–e; Table S3). The SMGY model offers a user-friendly, efficient, and non-destructive method for accurate Chl quantification, facilitating rapid monitoring of Chl fluctuations while preserving plant integrity. To address potential false positives from NbSGR1 gene activation in N. benthamiana, we used CRISPR/Cas9 technology to knock out its six SGR1 homologous genes, distributed across different chromosomes (Figure S5a). We designed a CRISPR-Cas9 construct with 12 sgRNAs under the AtU6-26 promoter, which was introduced into N. benthamiana (Figure S5b). Genetic analysis revealed a homozygous plant, CR19, with all six NbSGR1 genes edited (Figure S6a). CR19 showed delayed leaf yellowing and reduced Chl degradation under dark conditions, making it ideal as a host for subsequent SRS research (Figures 1e–h and S6b–d). To increase the likelihood of TFs and target DNA interacting within the same cell, we developed the pTF-SGR1 plasmid featuring two independent expression cassettes: p35S::TF and pY::SGR1, with multiple cloning sites for ease of molecular manipulation (Figure S7). We evaluated this system using the FAR1 TF and the FHY1 promoter, which regulates the nuclear accumulation of phytochrome A. Interactions were confirmed via ChIP-PCR, Y1H, and EMSA (Lin et al., 2007). We constructed pFHY1::SGR1-p35S::FAR1, infiltrated N. benthamiana leaves with it, and observed significant colour shifts from green to yellow, indicative of interaction, while controls showed minimal changes (Figures 1i–p and S8). Elevated expression of SGR1 in the presence of FAR1 and the FHY1 promoter was confirmed (Figure 1q–t). Specificity tests with a mutated FAR1 gene and a non-interacting OsTB1 promoter validated the system's sensitivity and specificity (Figure S9a,b). Negative controls included vectors with either the FAR1 gene or the FHY1 promoter alone, and an empty vector, with minimal changes observed in controls (Figure S9c–f). To assess the SRS's ability to characterize interactions between TFs and their target promoters across diverse functions, we tested three TF-promoter pairs (Figure S10). For example, the TIG1 TF from the TCP family, which activates SAUR39 and influences rice tiller angles (Zhang et al., 2019), was tested by delivering the pSAUR39::SGR1-p35S::TIG1 vector into N. benthamiana leaves. This resulted in significant colour changes and increased SGR1 transcript levels, unlike controls (Figures S10, S11a–e and S12a,b). We also examined the interaction between MYB29 TF and the SUR1 promoter (Ma et al., 2013), observing expected colour changes with the pSUR1::SGR1-p35S::MYB29 construct (Figures S10, S11f–j and S12c,d). Additionally, the interaction between the avrBs3 protein from Xanthomonas campestris and the pepper Bs3 promoter (Römer et al., 2007) was confirmed through noticeable colour changes upon co-expression (Figures S10, S11k–o and S12e,f). Furthermore, we demonstrated the feasibility of the SRS system in validating the interaction between the senescence-associated TF AtNAP (Zhang and Gan, 2012) and its downstream target gene, the SAG113 promoter (Figure S13). These experiments demonstrate the SRS's robustness and reliability in verifying specific plant TF–DNA interactions. We further tested the SRS in various plants beyond N. benthamiana, confirming its effectiveness in species like rapeseed and different types of lettuce (Figure S14). However, some plants showed unexpected phenotypes, indicating the need for future optimization of experimental conditions. This work as supported by Biological Breeding-Major Projects (2023ZD04076), National Key Research and Development Program of China (2023YFF1000103), National Natural Science Foundation of China (32072115), and the CAAS Innovation Program to C.L. The authors declare no competing interests. CL designed the research; QZ, HH, ZM, TZ, YaL, JiZ, ZL, and YW conducted experiments; QZ, YY, JingZ, and SG analysed data. CL and QZ wrote the manuscript. Data supporting this study's findings are in the supplementary material. Figure S1 Phylogenetic examination of SGR1 TF homologs in plants, with color-coded subgroups. Figure S2 Constructs for SGR1s corresponding to Figure 1a. Figure S3 Dark treatment accelerated yellowing of SGR1-injected leaf locations. Figure S4 Development of SMGY calculation model. Figure S5 NbSGR1 genes sequence and CRISPR/Cas9 vector construction. Figure S6 Editing of six NbSGR1 homologs utilizing CRISPR/Cas9 in N. benthamiana for SRS. Figure S7 Schematic of construct, pTF-SGR1. Figure S8 Schematic diagrams of constructs for confirmation of the interaction between the FAR1 TF and the FHY1 promoter. Figure S9 Evaluation of the specificity of the SRS through examination of the interaction between the FAR1 TF and the corresponding gene promoter. Figure S10 Schematic diagrams of constructs used for verification of the interaction between the TIG1-SAUR39 promoter, MYB29-SUR1 promoter, and avrBs3-Bs3 promoter. Figure S11 Three different TFs and their respective gene promoter interactions revealed by the SRS. Figure S12 Gene expression analysis corresponding to Figure S11. Figure S13 AtNAP TF and SAG113 promoter interactions revealed by the SRS. Figure S14 SRS assessment in diverse species. Table S1 Fourteen features significantly correlated with SPAD values. Table S2 Modelling iteration results of five machine learning predictive models and one stacking ensemble learning model. Table S3 Performance metrics of the stacking ensemble model for SPAD value prediction. Table S4 Primers used in this study. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.