Network analysis reveals that bHLH transcription factors positively regulate rice resistance to Magnaporthe oryzae

Abstract

Transcription factors (TFs) and nucleotide-binding leucine-rich repeat (NLR) genes are core components of the immune response in rice against Magnaporthe oryzae. Although the construction of the rice pan-genome has provided new opportunities for dissecting the regulatory relationship between them, researchers still have limited systematic understanding of their interaction network during pathogen infection. In this study, we explored the interaction network between TFs and NLR genes by integrating the rice pan-genome with time-series transcriptomic data obtained at multiple time points during the rice-M. oryzae interaction. By combining weighted gene co-expression network analysis and the machine learning-based GENIE3 algorithm to construct TF-NLR interaction networks, we found that the bHLH TF family actively responds to M. oryzae infection and is widely involved in diverse biotic stress responses as well as the regulation of rice growth and development. To validate their biological functions, mutant rice lines were generated. Functional assays demonstrated that knockout of Os08g0490000, Os02g0726700, and Os01g0218100 significantly reduces rice resistance to M. oryzae infection (enhances susceptibility). Overall, this study highlights the importance of the bHLH TF family in rice immune responses through network-based analyses and identifies Os08g0490000 and Os01g0218100 as novel candidate genes associated with rice blast resistance. This study presents the first construction of a TF–NLR interaction network associated with rice blast resistance based on the rice pan-genome, providing new insights into the molecular mechanisms underlying rice–pathogen interactions. Furthermore, two novel bHLH-type resistance genes were identified, offering valuable genetic resources for rice improvement.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12870-026-09237-8.

Introduction

Rice (Oryza sativa L.) is a cornerstone crop for global food security, yet its production stability is continuously threatened by a wide range of biotic and abiotic stresses throughout the entire growth period [1, 2]. Among these challenges, rice blast disease caused by the fungal pathogen M. oryzae is one of the most destructive diseases, leading to severe yield losses and substantial economic damage. Consequently, rice blast has long been a major focus of plant pathology and crop disease resistance research [1, 3].

Current strategies for rice blast management mainly include chemical control [4], biological control [5], and the development and deployment of resistant cultivars [6], among which genetic resistance is regarded as the most economical and environmentally sustainable long-term solution.

Resistance (R) gene–mediated immunity represents a central theme in rice blast research. Since the first rice blast resistance locus was reported in the 1960s [7], more than 100 R genes or resistance loci have been identified to date [8]. Among them, 31 genes have been successfully cloned, beginning with Pib [6, 9], and several genes or locus members, such as Piz, Pik, and Pi-ta, have been widely applied in rice molecular breeding programs. In addition to NLR-type R genes that directly mediate disease resistance, accumulating evidence indicates that TFs play indispensable roles in the regulation of rice immune networks. For example, the Bsr-d1 allele confers broad-spectrum resistance to rice blast [10]; APIP5 regulates defense responses at both transcriptional and post-transcriptional levels [11]; and Pb1 functions through its CC domain by interacting with the TF OsWRKY45 [12, 13].

Recent studies have further demonstrated that TFs can directly target NLR genes, thereby fine-tuning plant immune responses to bacterial and fungal pathogens [14, 15].

With the expansion of rice genomics research from a single reference genome [16] to the pan-genome level [17–19], integrating multi-omics datasets to systematically dissect TF–NLR interaction networks and elucidate their roles in disease resistance has emerged as a critical research priority.

bHLH family plays vital roles in plant growth, development, and stress responses. In recent years, its function in regulating resistance against fungal diseases has been progressively elucidated. For instance, MdbHLH093 enhances resistance to powdery mildew in apple [20], while SlbHLH95 improves resistance to gray mold in tomato by modulating flavonoid metabolism [21]. In rice, bHLH25 confers broad-spectrum resistance by sensing H₂O₂ and regulating the lignin and phytoalexin pathways [22]; OsHLH46 interacts with the negative regulator OsbHLH6 to enhance blast resistance [23]; and OsbHLH5, OsbHLH057, among others, also play important roles [24, 25].

Nevertheless, the number of bHLH genes identified as regulators of blast resistance remains limited, with the majority of their functions still unknown. Moreover, existing studies have largely focused on individual genes, and systematic dissection of the regulatory network remains insufficient. Therefore, systematic identification of bHLH TFs involved in resistance against Magnaporthe oryzae, elucidation of their molecular networks, and mining of superior alleles are key to understanding the association between the bHLH family and rice blast resistance.

With the in-depth research on rice pan-genomics, analytical methods based on multiple reference genomes have provided new opportunities for mining exploitable rice genes and laid a solid foundation for constructing transcriptional regulatory networks [18].

Relying solely on a single reference genome has made it difficult to comprehensively resolve genetic variations among different rice germplasms [26]. To this end, the pan-genome strategy has been proposed and applied. Compared with studies based on a single reference genome, pan-genomics can more efficiently identify additional presence–absence variations (PAVs) [27]. These variations are often closely associated with various traits in rice, including disease resistance [19], and may influence the regulatory relationships between TFs (TFs) and nucleotide-binding leucine-rich repeat (NLR) genes. Notably, pan-genomics can identify disease resistance-related genes that are missing or incorrectly annotated in the reference genome [28]. Therefore, constructing a TF-NLR interaction network using pan-genome data can not only rapidly identify disease-resistant haplotypes but also more accurately and intuitively reveal genetic variations and their impacts on regulatory networks [29].

In this study, we integrated a high-quality rice pan-genome dataset comprising approximately 3,000 accessions with time-series RNA-seq transcriptomic data collected at four time points following M. oryzae infection in two rice cultivars Nipponbare (Oryza sativa L. ssp. japonica cv.) and Zhonghua11(Oryza sativa L. ssp. japonica cv.). By systematically constructing interaction networks between TFs and NLR genes, we successfully validated multiple previously reported disease resistance–related genes and TFs. Further analyses revealed that the basic helix–loop–helix (bHLH) TF family may play a central regulatory role in rice responses to rice blast stress, and a set of bHLH family members exhibiting significant differential expression upon M. oryzae infection was identified. Together, these findings provide new insights into the molecular mechanisms by which bHLH TFs mediate rice disease resistance through NLR gene networks.

Materials and methods

Data collection

From the RPAN database, the rice pan-genome of around 3000 rice genomes was retrieved [17]. Using gffread (v0.11.7) [30] with the default settings, the CDS sequences were extracted from the genome sequences and the matching GFF file. As stated by Yang et al. (2021), the RNA-Seq files were retrieved from the NCBI SRA database using the accession number PRJNA661210. From the PlantTFDB database, we downloaded the transcription factor list for O. sativa subsp. japonica. The genome annotation file of MH63 [31], which was retrieved from the RIGW database (http://rice.hzau.edu.cn/rice_rs3/), was used to match the genes’ identifiers from the PlantTFDB database to the pan-genome sequences’ ID. With parameters of -evalue 1e-5, we compared the protein sequences of the unmatched transcription factors to the pan-genome protein sequences and chose the candidate gene with the highest similarity for each unmatched gene.

Identification of NLR genes

The protein sequences of the pan-genome were utilized to identify NLR genes using two pipelines: NLRAnnotator (https://github.com/steuernb/NLR-Annotator) [32] and HRP (https://github.com/AndolfoG/HRP) [33]. The CDS sequences of all genes were utilized using default settings for NLRAnnotator. The first step of HRP entails utilizing InterProScan 5 [34] and the options -f TSV -app Pfam to search the entire NB-ARC domain. The discovered genes were divided into full-length and partial-length NLR genes.

Using genBlastG (v1.0.138) [35] with the default settings, the full-length NLR genes were then utilized to de novo blast novel NLRs gene(s) from the pan-genome. The common NLR genes discovered by the two approaches were chosen for this study as potential NLR genes.

RNA-Seq data processing

Parallel-fastq-dump (v0.6.7) (https://github.com/rvalieris/parallel-fastq-dump) was used to separate the SRA data that was downloaded from the NCBI SRA database into fastq-type files. All fastq-type files had their data quality checked using FastQC (v0.11.9).

The pan-genome index was created using HISAT2 (v 2.2.1) [36], and the RNA-Seq data were then automatically aligned to the pan-genome data using default settings. After sorting with samtools (v1.6) [37], the aligned SAM format file was converted to a BAM format file. Using featureCounts (v2.0.2) [38] and the options -p -t exon -g gene id, the counts value and the gene length were determined. Finally, the FPKM and TPM expression values were computed using our custom R script.

Differential expression analysis

The framework of the DESeq2 (v1.34.0) [39] was utilized to find the differentially expressed genes of RNA-Seq data with mapped read counts generated using featureCounts (v2.0.2) [38], as described by SI Zandalinas, S Sengupta, FB Fritschi, RK Azad, R Nechushtai and R Mittler [40]. To identify genes that were significantly differentially expressed, a cutoff of absolute log2(fold change) > 1 and the corresponding adjusted P-value ≤ 0.05 was used.

Weighted correlation network construction

Weighted correlation network analysis was carried out manually using the R package WGCNA (v1.72.1) with default settings unless otherwise noted by Langfelder and Horvath (2008) to create the network between transcription factors and NLR genes. In the WGCNA analysis, genes with counts > 5 and detection in at least 10% of samples were retained [41]. The default parameters of the R function goodSamplesGenes from the WGCNA package were used to pre-check the quality of both the samples and the genes. A signed network was employed to capture the optimal pattern of co-regulation, as described by Langfelder (2008). Using the pickSoftThreshold function, it was discovered that a soft threshold of 9 best fits the network structure. To isolate relatively large gene modules, we choose the minimum module size of 30. Using biweight midcorrelation, module eigengenes were then linked to sample metadata, such as species, inoculation status, and inoculation time. Using the R package linkET (v0.0.3) (https://github.com/Hy4m/linkET), the correlation between modules and between modules and sample meta information was shown. Gephi (v0.10.1) was used to display the weighted correlation networks. Otherwise, default parameters were utilized.

GENIE3 network

To filter out low-expression noise, only genes with an expression level of FPKM/TPM > 1 in at least 10% of the total RNA-seq samples were retained for subsequent network analysis. In order to estimate the random forest regression for each gene using default settings, 1862 TFs were entered into the R program genie3 (v1.16.0).

Default parameters were used (K=sqrt, n.trees = 1000, regulators = list of TF, importance.measure = IncNodePurity, seed = NULL). All predicted target NLR genes (connectivity > 0.005) [42] for each TF were gathered and then utilized the network out degree > 21 (the median value of all TFs) to filter the network. Gephi was finally used to visualize the GENIE3 networks (v0.10.1).

Expression of bHLH transcription factor members

In order to investigate the expression of bHLH TF members, we obtained an extensive amount of transcriptome data from the NCBI SRA database. This data was collected from 23 different projects, which covered six different biotic stresses (Supplementary Table 1). These stresses include Burkholderia vietnamiensis, M. oryzae, Ustilaginoidea virens, Xanthomonas oryzae, Herbaspirillum seropedicae, and Nematode.

The raw RNA-Seq data underwent processing procedures as described earlier. The fold-change value for 9 common bHLH TF members, which were selected using Venn analysis, was calculated using our custom R script. The data was then visualized through the ggplot2 package (v3.4.1).

Rice blast inoculation and qPCR analysis

To validate the functions of seven genes identified from RNA sequencing (RNA-seq) expression data, we cultivated the rice variety ‘Nipponbare’ (O. sativa. ssp. japonica cv.) under indoor conditions. For each treatment group, six pots serving as independent biological replicates were established, with five plants planted per pot. When the plants reached the “three-leaf and one-heart” growth stage, three of these pots were selected for inoculation experiments with rice blast fungus. A conidial suspension of the M. oryzae strain Guy11 (5 × 10⁴ spores mL⁻¹) was evenly sprayed onto the leaf surfaces of rice plants. After inoculation, plants were incubated in the dark at 26 °C with 95% relative humidity for 16 h, followed by growth under controlled conditions with a 28 °C /25°C day/night temperature regime and a 14 h/10 h light/dark photoperiod for 7 days, until typical blast lesions developed.

Leaf tissues from infected plants were collected as treatment samples, while leaves from non-inoculated healthy plants were harvested as controls. Total RNA was extracted from rice leaf tissues using a Plant RNA Extraction Kit (TIANGEN, China), and RNA concentration and purity were assessed using a NanoDrop 2000 spectrophotometer.

Subsequently, first-strand cDNA was synthesized from total RNA using the Fast Quant RT Super Mix reverse transcription kit (TransGene, Beijing, China) according to the manufacturer’s instructions.

Quantitative real-time PCR (qRT-PCR) was performed using the Real Master Mix (SYBR Green, TransGene, Beijing, China) on an ABI 7500 Real-Time PCR System (Applied Biosystems, USA). The PCR cycling conditions were as follows: an initial denaturation at 95 °C for 2 min, followed by 40 cycles of denaturation at 95 °C for 15 s, annealing at 60 °C for 25 s, and extension at 72 °C for 30 s. The rice gene LOC_Os01g22490 was used as the internal reference gene for normalization. Relative gene expression levels were calculated using the 2^⁻ΔΔCt method based on Ct values obtained at a defined fluorescence threshold.

Mutant generation and functional validation

Candidate genes that showed significant associations with disease resistance in the above analyses were selected for functional validation. In the genetic background of Nipponbare, highly specific candidate gRNA sequences were designed using the CRISPR-GE tool. These sequences were further filtered by local BLAST analysis to ensure that the target sites were located near the 5′end of the coding sequence (CDS) and exhibited minimal off-target effects [43]. Based on the selected target sequences, forward and reverse oligonucleotides were synthesized and annealed to form double-stranded DNA fragments. These fragments were then directionally inserted into the expression vector using the Golden Gate cloning system to generate recombinant plasmids. The constructed plasmids were subsequently introduced into Agrobacterium tumefaciens strain EHA105 via electroporation. After induction, rice calli were co-cultivated with the transformed Agrobacterium for two days, followed by transfer to a hygromycin-containing selection medium for screening. The resistant calli were then transferred to a rooting medium for regeneration. Once regenerated into seedlings, genomic DNA was extracted and PCR analysis was performed to identify positive transgenic plants. Functional validation was primarily conducted using homozygous mutants in the T₁ generation. To eliminate potential effects of transgenic components on phenotypes, Cas9-free lines were further screened in the T₁ generation by PCR. Finally, Sanger sequencing was performed to confirm the homozygous mutation status at the target sites. To further evaluate the role of the mutant plants in rice blast resistance, wild-type plants and homozygous mutant plants (T₁ generation) at the three-leaf–one-heart stage were subjected to M. oryzae inoculation. The experiment set up three independent biological replicates, each containing 6 plants. A conidial suspension of the M. oryzae strain Guy11 (5 × 10⁴ spores mL⁻¹), prepared in 0.1% (w/v) gelatin, was evenly sprayed onto the leaf surfaces of rice plants. Following inoculation, plants were incubated in the dark at 26 °C and 95% relative humidity for 16 h to facilitate pathogen infection.

Subsequently, plants were transferred to controlled growth conditions with a 28 °C/25°C day/night temperature regime and a 14 h/10 h light/dark photoperiod for 7 days until typical rice blast lesions developed. The degree of susceptibility or resistance in rice is assessed based on the number of lesions per unit area.

Results

The pan-genome outperforms a single reference genome in NLR gene identification

In this study, two complementary approaches, NLRAnnotator and HRP, were employed to identify nucleotide-binding leucine-rich repeat receptor (NLR) genes across a rice pan-genome comprising approximately 3,000 accessions. Using NLRAnnotator and HRP, we identified 503 and 937 NLR genes, respectively (Fig. 1A, B). These numbers are substantially higher than those reported in previous studies based on a single reference genome [44, 45], highlighting the advantage of the pan-genome in capturing the diversity of large gene families.

Fig. 1.

NLR genes identified using two widely used pipelines. A The NLR genes identified by NLRAnnotator. B The NLR genes identified by HRP pipeline. The detail of each NLRs type can be found in the corresponding references

Notably, in addition to full-length NLR genes, HRP was able to identify truncated NLR variants. Given that truncated NLRs have been demonstrated to play functional roles in disease resistance [46], the results obtained from both methods were integrated. In total, 977 putative NLR genes were retained for subsequent analyses (Supplementary Table 1).

WGCNA reveals co-expression patterns between transcription factors and NLR genes

Weighted gene co-expression network analysis (WGCNA) was employed to systematically identify infection-specific co-expression modules in two rice cultivars during M. oryzae inoculation. High-quality genes and samples filtered using the ‘goodSamplesGenes’ function were used to construct a weighted network with a soft-thresholding power of β = 9, resulting in 68 distinct modules (Fig. 2A), with module sizes ranging from 35 to 5,369 genes and an average of 482 genes per module.

Fig. 2.

WGCNA-based identification of key modules and genes. A Correlation between WGCNA modules and treatments. Color intensity indicates the strength of correlation. Significance levels are indicated as P < 0.05 (*), P < 0.01 (**), and P < 0.001 (***). B Venn diagram showing modules significantly associated with different time points post-inoculation. C Distribution of transcription factor families and their abundances within WGCNA modules. D Gene expression profiles of genes in the yellow module. E Interaction network between transcription factors and NLR genes constructed from significant modules

To investigate associations between modules and traits, a trait matrix was generated incorporating cultivar, inoculation treatment (Guy11 strain), and time-point information using the model.matrix function. While most modules exhibited weak correlations with the traits, several modules showed significant associations (Fig. 2A). Notably, the yellow module displayed opposing correlation patterns between the two cultivars.

Analysis of the expression patterns within the yellow module revealed that the genes in this module exhibit a time-dependent fluctuating expression pattern and lack variety specificity (Supplementary Fig. 1). Among these, the rhythmic gene Os07g0257200 (MH07g0175300), known to respond to elevated night-time temperatures and regulate environmental adaptation [47], suggests that approximately 2,300 rhythmic genes may integrate environmental signals through similar mechanisms. The yellow module contained 25 NLR genes and 129 TFs from 33 families, indicating that circadian transcriptional regulation may contribute to the temporal expression of specific NLR genes.

To focus on infection-responsive core modules, modules significantly associated with each time point post-inoculation were identified. Seventeen, seven, seventeen, and twenty-seven modules were significantly associated at 12, 24, 36, and 48 h post-inoculation, respectively (Fig. 2B), while no uniquely responsive modules were observed at 48 h. TFs and NLR genes from all significant modules were extracted using the exportNetworkToCytoscape function (weight threshold = 0.01, weighted mode) to construct a global post-inoculation regulatory network. This network revealed that members of the bHLH, ERF, MYB, WRKY, NAC, bZIP, and C2H2 families formed the majority of network nodes (Fig. 2C, Supplementary Table 2), consistent with their previously reported central roles in rice disease resistance [8].

Based on these results, a weighted TF – NLR interaction network was further constructed to delineate interaction patterns (Fig. 2E). The number of TFs varied across modules; for example, the brown module contained 109 TFs and was significantly associated with six trait groups, indicating that M. oryzae infection markedly reshapes the co-expression network between TFs and NLR genes (Fig. 2D).

Multiple putative transcription factors involved in rice blast response revealed by GENIE3 network analysis

GENIE3 network analysis identified 31 TF families that may regulate NLR genes at different infection stages in multiple rice cultivars (Table 1). Among these, the functions of eight families have been validated in previous studies, including HSF [48, 49], WRKY [50–52], NAC [53, 54], SBP [48], C2H2 [10, 55], MYB [56–58], ERF [59, 60], and LSD [61]. Notably, the network included the broad-spectrum resistance gene BSR-D1/ZFP36 (Os03g0437200), which is predicted to target 44 NLR genes [62]. The remaining 23 TF families represent previously unreported regulatory components, highlighting the complexity of rice blast resistance mechanisms.

Table 1.

Identified transcription factors related to the response to rice blast infection

TFs family	Mean connectivity	Co-expressed NLrs numbers
bHLH	0.0139	1012
WRKY	0.0129	792
NAC	0.0136	513
MYB	0.0120	450
MYB_related	0.0128	439
C3H	0.0131	415
ERF	0.0127	411
C2H2	0.0115	361
HSF	0.0136	334
G2-like	0.0117	244
DBB	0.0132	213
Dof	0.0160	208
CO-like	0.0124	208
GRAS	0.0134	200
HD-ZIP	0.0128	188
bZIP	0.0126	158
ARF	0.0109	119
ZF-HD	0.0128	105
B3	0.0193	100
Trihelix	0.0140	74
SBP	0.0134	60
LBD	0.0104	57
Nin-like	0.0120	56
GeBP	0.0150	54
RAV	0.0109	50
M-type_MADS	0.0119	50
HB-other	0.0117	50
LSD	0.0131	50
EIL	0.0108	48
BES1	0.0127	43
E2F/DP	0.0107	42

Network analysis revealed that these eight TF families exhibited specific connections with different NLR genes (Fig. 3A). Among 25 core hub genes, 12 belonged to the WRKY family, further confirming the central role of WRKY TFs in rice blast resistance [63]. Heatmap analysis showed that only a subset of TFs were co-expressed with NLR genes during infection (Fig. 3B), consistent with studies demonstrating that CC-NBS-LRR R proteins transmit defense signals via WRKY TFs [64].

Fig. 3.

The GENIE3 network and heatmap of reported rice blast related-transcription factors and the bHLH transcription family. A The GENIE3 network of previously studied rice blast-related transcription factor families. Transcription factors were labeled with their gene names used in related studies and the green color represents the co-expressed NLR genes. B The heatmap of reported rice blast related transcription factors and corresponding co-expressed NLR genes. The expression value was scaled using a function from the R package pheatmap (v1.0.12). C The GENIE3 network of the bHLH transcription factor family and its co-expressed NLR genes. The bHLH transcription family was colored in green with its gene name in the pan-genome and the red color is the NLR genes. D The heatmap of the bHLH transcription factor family and corresponding NLR genes. The expression value was scaled using a function from the R package pheatmap (v1.0.12)

Importantly, the bHLH family was predicted to regulate the largest number of NLR genes (Table 1). In the present network, 16 bHLH TFs were identified (Fig. 3C).

Although their direct involvement in rice blast resistance has not been previously reported, the significance of bHLH TFs in plant immunity has been documented: for example, OsbHLH6 mediates resistance by dynamically regulating the JA/SA signaling pathways [65], and RAI1 functions downstream of OsRac1 to modulate immune responses [66]. Additionally, OsDPF is involved in the regulation of diterpenoid phytoalexin biosynthesis [67]. Expression analysis further revealed that these bHLH TFs exhibit cultivar-specific patterns and participate in the plant’s disease resistance response 12 h after inoculation, with the strongest response observed in the Nipponbare cultivar (Fig. 3D).

bHLH transcription factor family members respond to multiple stresses

Both WGCNA and GENIE3 network analyses highlighted the central role of the bHLH TF family in rice immune responses; however, the breadth of their functional roles remains unclear. Given the pleiotropic nature of TFs (e.g., OsWRKY71 regulates both defense and cold stress responses) [68, 69], we integrated data from 23 NCBI SRA projects to systematically investigate the biological functions of this family.

Among the seven bHLH genes identified at the intersection of the two networks (two of the original nine candidate genes were missing), all but Os06g0570900 were upregulated under multiple biotic stress conditions (Fig. 4). The results of the significance test are shown in Supplementary Table 3. Importantly, five bHLH members responding to M. oryzae were identified for the first time. Notably, Os02g0726700 and Os02g0603600 were previously reported to respond to X. oryzae pv. oryzae, the causal agent of bacterial leaf blight [70, 71], highlighting their potential roles in cross-pathogen regulation.

Fig. 4.

The fold-change value of bHLH transcription factor family members induced by various biotic stresses. In each panel, the dashed line on the left represents fold change (log2) = -1, and the dashed line on the right represents fold change (log2) = 1

Of the 2,063 genes obtained from the metadata combination of 23 datasets, these genes were significantly induced by rice blast infection (log₂FC > 1), while only 108 genes showed downregulation, confirming their predominantly positive regulatory function. Moreover, beneficial rhizosphere bacteria, including B. vietnamiensis and Azospirillum, were found to modulate the expression of most bHLH members, suggesting novel functions in nitrogen fixation and nutrient uptake [72–75]. In addition, differential expression under Fusarium fujikuroi and nematode infections further expands the functional spectrum of this family, indicating their involvement in broad-spectrum disease resistance.

Validation of candidate network genes by qRT-PCR and CRISPR-Cas9 gene editing

To validate the expression of the seven candidate genes identified at the intersection of the networks, wild-type rice plants grown under normal conditions were used as controls and inoculated with M. oryzae. qRT-PCR analysis revealed that all seven intersection genes were upregulated in infected plants compared with uninfected wild-type plants. Among these, Os08g0490000, Os02g0726700, and Os01g0218100 showed significant induction upon pathogen infection (Fig. 5A).

Fig. 5.

Expression and functional validation of bHLH transcription factor family members under M. oryzae infection in rice. A Expression levels of candidate genes in Nipponbare leaves after inoculation with M. oryzae, with healthy plants used as controls. B Genotypic characterization of the three knockout lines. C Lesion number and disease phenotypes of gene-edited knockout lines following M. oryzae inoculation, with wild-type plants as controls. Differences among groups were analyzed using one-way ANOVA followed by Tukey’s multiple comparison test. P < 0.05 (*), P < 0.01 (**)

To further investigate the roles of these three candidate genes in rice blast resistance, CRISPR-Cas9 gene editing was used to generate knockout lines for Os08g0490000, Os02g0726700, and Os01g0218100. Specific sgRNAs were designed to target the coding regions of the candidate genes for genome editing. Sanger sequencing of the transformants successfully identified multiple homozygous mutant lines carrying targeted edits. As shown in Fig. 5B, the os01g0218100-KO6 line harbors a 1 bp deletion accompanied by a base substitution; the os02g0726700-KO7 line contains a 4 bp deletion along with a base substitution; and the os08g0490000-KO16 line exhibits a 1 bp deletion.

These insertion/deletion (InDel) mutations result in frameshift changes and premature termination of translation, thereby leading to complete loss of function of these bHLH TFs. Phenotypic analysis revealed that, compared with wild-type plants, all gene-edited lines exhibited a significant increase in lesion number and lesion density upon M. oryzae infection, indicating enhanced susceptibility (Fig. 5C). These results clearly demonstrate that Os08g0490000, Os02g0726700, and Os01g0218100 play essential roles in bHLH-mediated rice blast resistance, positively regulating the defense response.

Importantly, Os08g0490000 and Os01g0218100 represent previously unreported rice blast resistance genes. This finding not only highlights the utility of our network-based analysis for identifying novel resistance components but also provides valuable genetic resources for molecular breeding of rice with enhanced blast resistance.

Discussion

Plant immunity relies on the molecular recognition mediated by immune receptors.

Although effector proteins secreted by M. oryzae during invasion can induce susceptibility, they are frequently recognized by NLR (nucleotide-binding site–leucine-rich repeat) receptors, triggering effector-triggered immunity (ETI) and thereby conferring resistance to rice [76]. As a core component of plant immunity, functional characterization of NLRs is essential for elucidating rice disease resistance mechanisms [77]. However, the NLR family is large and structurally highly variable, comprising an N-terminal domain, a central NB-ARC domain, and a C-terminal LRR domain, with extensive diversity at the sequence, structural, and regulatory levels [78].

Notably, some truncated NLRs retain resistance functions, making it difficult to capture the full spectrum of NLR variation using a single reference genome [79]. In this study, we integrated a rice pan-genome comprising 3,000 accessions and employed NLRAnnotator and HRP tools to identify 977 NLR genes, highlighting the advantages of pan-genomic analysis in capturing both genetic diversity and structurally variable NLRs.

In this study, a key finding from the WGCNA analysis was the identification of the “yellow module,” which contains 25 NLR genes and 129 TFs. The opposite correlation patterns observed between resistant and susceptible varieties suggest that infection by M. oryzae may suppress host defense by interfering with the host circadian clock. Although our sampling at four time points (12, 24, 36, and 48 h post-inoculation, hpi) provides only a preliminary view of these dynamics, the results indicate that the circadian system may regulate the expression of specific NLR genes to optimize resource allocation between growth and immunity. Future studies with higher temporal resolution are required to confirm the precise rhythmic oscillations of these components during the infection process.

Immune TFs enhance ETI efficiency by activating specific defense genes, and their direct interactions with NLRs can significantly increase defense effectiveness [80].WRKY, NAC, and bHLH are major TF families involved in plant immunity, playing crucial roles in activating pathogenesis-related (PR) genes, triggering the hypersensitive response (HR), and establishing systemic acquired resistance (SAR). Acting as central regulatory hubs, these TFs integrate signaling molecules such as salicylic acid (SA), jasmonic acid (JA), and ethylene (ET), thereby enabling precise spatiotemporal control of defense responses [81]. Among them, WRKY represents one of the best-characterized and functionally defined regulators of NLR genes. WRKY proteins can directly bind to W-box elements in the promoters of NLR genes, and have also been shown to interact with the CC domains of NLR proteins such as Pi36, Pib, Pita, Pit, and Piz-t, thereby regulating downstream defense gene expression [64]. So, WRKY can activate the transcriptional expression of NLR genes and is thus regarded as a direct upstream regulator of NLR gene expression, playing a central role in the rapid activation of effector-triggered immunity [82]. NAC TFs regulate gene expression through diverse mechanisms and participate in the integration of stress signals and developmental signals, forming flexible and complex regulatory networks that consequently establish a favorable cellular environment for NLR-mediated resistance responses [83]. Although WRKY and NAC families are widely recognized as “early responders” that can directly bind cis-regulatory elements in NLR promoters, our GENIE3 network analysis identified the bHLH family as a potential master coordinator, showing associations with the largest number of candidate NLR target genes. Based on this, we propose a potential division-of-labor model: WRKY factors may primarily trigger rapid acute defense activation, whereas bHLH factors, acting as signal integrators, may fine-tune long-term systemic resistance by modulating hormone signaling pathways (JA/SA) and secondary metabolism. Interestingly, Os06g0570900 showed no detectable response under multiple stress conditions, suggesting a high degree of functional specificity. This gene may be dedicated to regulating specific developmental stages (such as floral organ development), or may respond only to particular abiotic stresses (such as heavy metals or extreme temperatures). Alternatively, it may function as a transcriptional repressor within the immune regulatory network, where its downregulation or maintenance of silence could be a prerequisite for activating downstream defense pathways in certain immune responses.

As the second largest TF family in plants, the bHLH characterized by a conserved E-box binding domain [84]. Members of the bHLH family participate in plant pattern-triggered immunity (PTI) and systemic acquired resistance (SAR) mainly by regulating hormone signaling pathways such as ET and SA, or by directly activating the expression of genes involved in phytoalexin biosynthesis [85]. However, their specific functions in the regulation of NLR genes remain to be further elucidated.

To systematically dissect this interaction network, we applied weighted gene co-expression network analysis (WGCNA) to cluster key genes under M. oryzae infection into 68 modules, revealing that TFs may regulate NLR gene expression in response to temperature changes. Based on these findings, we constructed a TF-NLR weighted correlation network, providing a foundation for further mechanistic investigation.

GENIE3 network analysis further predicted the regulatory connections between TFs and NLR genes [86]. We identified 31 TF families potentially regulating NLRs, eight of which have previously been validated in disease resistance, while 23 represent newly discovered components. Particularly notable is the bHLH family, comprising 16 members predicted to regulate 1,012 NLR genes. WGCNA analysis also revealed bHLH as the most enriched TF family, establishing its central role in rice blast resistance. Plant growth-promoting rhizobacteria (PGPR, such as B. vietnamiensis), can induce induced systemic resistance (ISR) in plants, a process that is typically highly dependent on jasmonic acid (JA) and ethylene (ET) signaling pathways. bHLH TFs, including the well-characterized MYC2, serve as central integrators within the JA signaling pathway and play pivotal roles in regulating defense responses. Therefore, the upregulation of these bHLH genes in response to beneficial rhizosphere microbes suggests that they likely function as key signal relay nodes in the systemic defense network linking “root microbial perception” to “activation of aboveground disease resistance.” Moreover, the ability of bHLH family members to respond to diverse pathogens further highlights their potential role as central hubs integrating multiple biotic stress signals.

By integrating bioinformatic approaches with large-scale RNA-Seq data, we identified nine candidate bHLH genes, among which Os08g0490000, Os02g0726700, and Os01g0218100 were confirmed to positively regulate rice blast resistance. Gene knockout experiments further validated two previously unreported resistance genes, providing valuable genetic resources for molecular breeding. Importantly, these genes exhibit multi-pathogen responsiveness: they are upregulated by bacterial leaf blight infection, modulated by beneficial nitrogen-fixing bacteria (B. vietnamiensis and Azospirillum), and show differential expression upon F. fujikuroi and nematode infection.

This multifunctionality is consistent with the role of OsbHLH96 in coordinating brown planthopper resistance and JA-SA signaling pathways, indicating that bHLH TFs serve as key regulatory hubs integrating responses to biotic and abiotic stresses [87].

The bHLH TF family is highly conserved across eukaryotes, characterized by a typical structure consisting of a basic region and a helix–loop–helix (HLH) domain, which are responsible for DNA binding and protein dimerization, respectively [88]. In rice, functionally characterized disease resistance-related bHLH genes all contain the canonical bHLH domain, and their ability to bind E-box elements is essential for activating the expression of downstream defense genes [89].The high structural conservation of bHLH proteins suggests that the candidate genes identified in this study may regulate downstream NLR gene expression through similar DNA-binding and dimerization mechanisms. However, their evolutionary relationships and functional conservation remain to be further elucidated through comprehensive comparative analyses.

In summary, this study provides a comprehensive dissection of the TF–NLR interaction network, establishing the bHLH family as a central regulatory module and identifying novel resistance genes. These findings offer valuable resources for elucidating rice blast resistance mechanisms. Although further experimental validation is required, in-depth functional and regulatory pathway analyses of these key genes may inform new strategies for molecular breeding, facilitating the development of broad-spectrum disease-resistant rice varieties and contributing to global food security.

Although this study highlights the important role of the bHLH TF family in rice resistance to M. oryzae, several limitations remain. First, we only preliminarily identified the disease-resistance functions of the relevant bHLH genes, and their regulatory targets and downstream signaling pathways have not yet been systematically characterized.

Whether these proteins directly bind to and regulate the promoters of NLR genes requires further validation using molecular interaction approaches such as yeast one-hybrid (Y1H) assays and chromatin immunoprecipitation (ChIP). Second, the downstream physiological mechanisms regulated by these TFs-such as reactive oxygen species (ROS) bursts, callose deposition, and the qPCR expression patterns of hormone-related marker genes-remain to be elucidated. Finally, although these candidate genes show promising application potential, their effectiveness in natural field conditions and across diverse genetic backgrounds must be rigorously evaluated to fully realize their value in molecular breeding. Future research should focus on several key aspects: elucidating the direct regulatory mechanisms between bHLH proteins and NLR genes through molecular interaction assays; validating the disease-resistance effects and breeding potential of candidate genes in diverse rice genetic backgrounds; conducting multi-location and multi-year field trials to systematically assess disease resistance and agronomic traits; and exploring the roles of bHLH TFs in responses to abiotic stresses to fully uncover their central functions in stress regulatory networks.

Conclusions

This study leveraged the advantages of the rice pan-genome to systematically capture genetic diversity and identify multiple candidate NLR genes. By integrating weighted gene co-expression network analysis (WGCNA) with GENIE3 network inference, we constructed a TF-NLR interaction network under M. oryzae infection, revealing the central role of the bHLH family in regulating NLR gene expression.

Subsequent RNA-Seq analysis identified M. oryzae-inducible bHLH genes, and CRISPR/Cas9-mediated functional validation confirmed their contribution to disease resistance, including the discovery of two previously unreported resistance genes, Os08g0490000 and Os01g0218100. Taken together, this study establishes a comprehensive research pipeline “pan-genome mining - network construction- functional validation” and revealed the interaction network between bHLH TFs and NLR genes, providing a theoretical foundation for elucidating their synergistic regulatory mechanism in resistance against M. oryzae, and also offering a scientific basis for resistance gene mining and germplasm innovation.

Authors' contributions

X.L. and H.H. designed the study; F.L., Y.L. and X.C. assisted with analysis design; L.M., Y.L. and F.L. collected the data; L.M. and Y.M. conducted data analysis and visual plotting; Y.M., Y.L.，Y.B. and X.C. interpreted the results; L.M., Y.M., F.L., X.L. wrote the manuscript, with final approval from all the co-authors.

Funding

This study was supported by National Key R&D Program of China (2023YFE0107500), National Natural Science Foundation of China (32460705), Yunnan Provincial Major Science and Technology Project (202402AE090026), Yunnan Fundamental Research Projects (202401AS070002) and Yunnan Xingdian Talent Support Program.

Data availability

All datasets used in this study are publicly available in the cited public databases. Details can be found in the Methods section. Processed data (e.g., gene expression matrices, network files, or lists of differentially expressed genes supporting network analysis) can be obtained from the corresponding author upon reasonable request.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.