BnSTINet: An experimentally‐based transcription factor interaction network in seeds of Brassica napus

For the past decade, research on constructing large‐scale plant protein–protein interaction (PPI) networks has mainly focused on model plants, and the capability of experimental‐based large‐scale PPI networks in these plants was thought to be significantly inferior to the actual number of protein interactions in plants (Altmann et al., 2020; Jones et al., 2014; Smakowska‐Luzan et al., 2018; Wang et al., 2023). Currently, there is still a lack of experimentally‐derived, large‐scale PPI data in plants, especially with respect to protein interaction networks specific to seed organs (Rhee and Mutwil, 2014). In this study, we selected Brassica napus, one of the most important oil crops extensively cultivated worldwide, to construct BnSTINet: an experimental‐based seed transcription factor (TF) interaction network.

CrY2H‐seq is a method that utilizes high‐throughput screening of PPIs and relies on precise protein expression libraries (Trigg et al., 2017). Nonetheless, the relatively high cost of this method restricts its extensive application (Pruneda‐Paz et al., 2014). We have innovated a method for constructing precise yeast expression vectors, referred to as In‐Gate, resulting in a remarkable cost reduction of 93% and a nearly 50% reduction in experimental time compared to the conventional CrY2H‐seq method (Figure S1). To screen the candidate TFs for constructing the interaction networks, the developing seeds at the linear embryo, early curved embryo, mid‐ curved embryo, late curved embryo and green embryo stages were collected for RNA‐seq (Figure 1a). The expression patterns of TFs were evaluated, and 10 gene clusters potentially related to seed development and organic accumulation were identified (Figures S2 and S3). Totally, a data set consisting of 932 non‐redundant TFs (NRTFs) was proposed to construct the TFs interaction network (Table S1). Almost all NRTFs with FPKMs value greater than 5 were included in the data set (99.1%) (Figure 1b). Those TFs were involved in seed development, such as hormone response, embryo development, seed coat development, lipid metabolism and light response, etc. (Table S2).

Figure 1.

The construction of TF interaction network in developing seeds of Brassica napus based on mCrY2H‐seq. (a) Sampling diagram of seed RNA‐seq in B. napus at different developmental stages. DAF, Day after flowering. (b) The coverage of candidate NRTFs for CrY2H‐seq interaction network construction. (c) Evaluation of the influence of self‐activated TFs on the screening of protein interaction. (d) Evaluation of interaction screening efficiency compared by nanopore and illuminate sequencing technology. ONT, oxford nanopore technologies. (e) The principle of screening protein interaction by CrY2H‐seq technology. (f) Workflow of mCrY2H‐seq in the PPIs screening. (g) Overview of TFs interaction network during seed development. (h) TF interactive network related to auxin signalling pathway; (i) Three‐step simplified BnSTINet information retrieval website user guide. (j) Geographical distribution of frameshift variant of BnaA03G0592400ZS in 2274 germplasm resources. (k) The influence of frameshift variant on oil content in BnaA03G0592400ZS gene of 258 rapeseed germplasm; ACT_Alt, the frameshift variant.

In total, a precise protein expression library consisting of 1886 Y2H strains expressing TFs was constructed one by one. Identification of self‐activating TFs and utilization of nanopore sequencing for detection of interacting recombinant sequences in TF interaction screening enhances the efficiency of detecting protein–protein interactions compared to the original CrY2H‐seq (Figure 1c,d). Based on the above optimization and modification of CrY2H‐seq technology, we term it mCrY2H‐seq which lowers the application threshold for the construction of experimental‐based large‐scale PPI networks (Figure 1e,f). We loaded a set of 1886 TFs into the mCrY2H‐seq pipeline. Self‐activating pre‐screen showed that 96 TFs were with self‐activating characteristics in the bait library (Figure S4; Table S3). Nanopore sequencing generated a total of 64 113 valid reads, while 62 747 of these reads were concurrently mapped to two different TFs, accounting for 97.9% of the entire set of valid reads. Totally, 805 pairs of PPIs were obtained and the unannotated TFs occupied 25.9% of total TFs in the network (Figure 1g; Table S4). The reliability of the interacting pairs within the TF interaction network was confirmed through one‐to‐one verification using the array yeast two‐hybrid method (Figures S5–S11). In the interaction network, multiple interaction subnetworks conserved in model plants were found, such as the Bzip family homodimers, nuclear factor complex, auxin and brassinosteroid (BR) signalling pathway TF interaction network (Figure S12). Some unreported interactions, such as the interaction between indole‐3‐acetic acid inducible protein 2 (BnaIAA2)/BnaIAA9, auxin response factor 6 (BnaARF6)/BnaIAA18 and BnaARF8/BnaIAA10 (Figure 1h). A total of 24 pairs of direct interactions and 54 pairs of indirect crosstalk were found between IAA, jasmonic acid (JA), abscisic acid (ABA), gibberellic acid (GA), BR and ethylene (ET) hormone signal pathways, which involved direct hormone crosstalk (Figure S13). The BnSTINet can be readily accessed and made available through a publicly accessible website with a simple three‐step process (https://yanglab.hzau.edu.cn/BnIR/TF_regulation_network) (Figures 1i and S14).

A candidate G‐box regulating factor 6 (BnaGRF6) that interacts with WRINKLED1 (BnaWRI1, a star molecule known to control fatty acid synthesis) was selected for the function investigation. The interaction between BnaGRF6 and BnaWRI1 was confirmed by Y2H, bimolecular fluorescence complementation (Figure S15a,b). Overexpression of BnaGRF6 significantly increased seed oil content by 0.9%–4.38% compared with the control (Figure S15c,d). Overexpression of the BnaGRF6 gene in seeds facilitated the elevation of transcriptional levels of downstream regulatory target genes of BnaWRI1 (Figure S15e). In 2274 B. napus germplasm in BnIR, a frameshift mutation from A to ACT in the reading frame of BnaGRF6 homologous gene BnaA03G0592400ZS was found (Figure S 16 a, b and c) (Yang et al., 2023). The frameshift mutation variant mainly appears in semi‐winter and spring rapeseed and is distributed in Asia, followed by Europe and North America (Figures 1j and S16d). Among 258 germplasm resources growing in the same environment, the average seed oil content of B. napus germplasm of the mutant variant was 41.74%, which was significantly lower than that of the reference type (43.42%) (Figure 1k). Here, we present a comprehensive research strategy for exploring the functions of polyploid crop genes and achieving molecular breeding advancements.

Conflict of interest

The authors declare no conflict of interest.

Author contributions

ML and YY designed the research. YY wrote the article. YY, ZG, JJ, HH, WC, NR, JL, YL, HZ and ZL optimized the CrY2H‐seq method, the expression vector construction, yeast mating and the recombinant plasmid information sequencing. YY, KC, HL, JH, WZ and YD collected and screened the transcript factors. ML, LZ, CF and YL revised the manuscript.

Supporting information

Figure S1 Workflow of the In‐Gate method.

Figure S2 and S3 NRTF clusters in B. napus seed.

Figure S4 Reads number of self‐activating TFs in the mCrY2H‐seq screen by nanopore sequencing.

Figures S5–S11 Point‐to‐point validation of interactions by array yeast two‐hybrid method.

Figure S12 Conserved subnetwork within BnSTINet.

Figure S13 The hormone crosstalk in BnSTINet.

Figure S14 A simplified tutorial on retrieving TF interactions in BnSTINet database.

Figure S15 Functional verification of BnaGRF6 affecting seed oil accumulation.

Figure S16 Variation and phenotypic distribution of BnaGRF6.

Table S1 Comprehensive TF list and expression data used for the BnSTINet construction.

Table S2 Enriched metabolic pathways list of the candidate NRTFs for mCrY2H‐seq interaction network construction.

Table S3 List of self‐activated transcription factors detected in the TF interaction network of seeds.

Table S4 List of all interaction pairs in the transcription factor interaction network of seeds.