The protein sequences of 33 rice accessions were downloaded from the Rice Resource Center (https://ricerc.sicau.edu.cn/), and the Hidden Markov Model (HMM) of the bZIP domain (PF00170) was downloaded from the Pfam database (http://pfam.xfam.org/). Using a threshold E-value of < 1e-5, HMMER (v3.3.2) was used to search the bZIP domain among protein sequence files of 33 rice accessions to extract candidate OsbZIPs. The protein sequences of candidate OsbZIPs were uploaded to SMART (http://smart.embl-heidelberg.de) to confirm the bZIP domain. Finally, according to the gene colinear relationships between 33 rice accessions in the study of Qin et al. [19], the genes that are colinear genes of the confirmed OsbZIPs were also regarded as bZIP gene family members.

The Arabidopsis bZIP protein sequences were obtained from the TAIR database (https://www.arabidopsis.org/). Multiple sequence alignments were conducted using MAFFT v7.490 with the default parameters. Subsequently, a phylogenetic tree was generated using FastTree version 2.1.11 [25] with the Jones–Taylor–Thornton (JTT) model and the maximum likelihood (ML) method and visualized through iTOL (https://itol.embl.de/upload.cgi).

The classification information of core and variable OsbZIPs was extracted from the study of Qin et al. [19] and a heatmap of variable genes constructed using R v4.0.3 ComplexHeatmap and RColorBrewer packages.

The Ka/Ks values of the colinear genes of each OsbZIP in 33 rice assemblies were calculated by KaKs Calculator Toolbox v2.0. The result was visualized using R v4.0.3 with ggridges and ggplot2 packages. The numbers and proportions of OsbZIP pairs with Ka/Ks >1 were counted. A heatmap was generated using R v4.0.3 ComplexHeatmap and RColorBrewer packages.

SVs that overlap with OsbZIPs were searched at the Rice Resource Center (https://ricerc.sicau.edu.cn/), and the expression values of genes with SVs and genes without SVs were derived from the study of Qin et al. [19]. The expression levels of OsbZIPs with and without SVs were plotted using GraphPad Prism v8.0.2.

The gff (general feature format) files were obtained from the Rice Resource Center (https://ricerc.sicau.edu.cn/). The protein sequences significantly affected by SVs were uploaded to MEME Suite v5.4.1 (https://meme-suite.org/meme/tools/meme), with the number of motifs set to 10. The resulting outcomes, along with the gff file of these proteins, were used as input files in TBtools (v1.098696) to obtain the gene structure.

The protein sequences of OsbZIPs in reference (90 OsbZIPs from NIP and 4 from CG14, which were lost in NIP) and OsbZIPs from FS32, which have the most OsbZIPs overlapped with SVs, were submitted to MEME Suite v5.4.1 (https://meme-suite.org/meme/tools/meme) with the number of motifs setting to 10 to obtain the sequence weblogos. The typical genes (containing the complete conserved domain) and atypical genes (not containing the complete conserved domain) in 33 accessions were plotted by ComplexHeatmap and RColorBrewer packages in R v4.03.

Raw RNA-seq data were downloaded from the NCBI Gene Expression Omnibus (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE79011). The fastq-dump of SRA Toolkit v2.10.0 was used to convert the SRA format files to FASTQ format. Fastp v0.20.1 was used to remove the adapter sequences and low-quality sequences. Clean reads were mapped to the reference genome (NIP) with HiSAT2 version 2.1.0 [26]. Fragments per kilobase of transcript per million mapped reads (FPKM) values of genes were calculated as the expression level. Differentially expressed genes were identified using DESeq2 version 1.32.0 [27] with a threshold |log2FoldChange| > 1 and FDR ≤ 0.05. A heatmap was generated with the transformed log2 (FPKM + 1) values using R (version 4.0.3) with the ComplexHeatmap package (version 2.6.2) [28].

Ninety-four OsbZIPs were identified by HMM search and SMART domains analysis (Table S1). These genes were named OsbZIP1-OsbZIP94 according to the order of their location. Among the 94 OsbZIPs, seventy-two OsbZIPs were core genes and twenty-two OsbZIPs were variable genes. Since the reference genome (NIP) does not contain all OsbZIPs, the phylogenetic tree was constructed with OsbZIPs from NIP (90 OsbZIPs) and CG14 (4 OsbZIPs which were lost in the reference genome) as well as AtbZIPs from Arabidopsis (Fig. 1A). According to the division of the AtbZIP phylogenetic tree in the previous study [20], Oryza sativa basic leucine zippers were divided into 11 groups. All the OsbZIPs in the group B, D, E, G, H, and I are core genes, suggesting that these groups were stable during domestication and improvement. Among the other five groups, there are the most variable genes in group S, indicating that the OsbZIPs in this group are more changeable in different accessions, which may be related to the unique traits of some accessions.

In the PAV analysis (Fig. 1B), groups A and C contain only one variable gene, the least number in those groups containing variable genes. These two genes are lost in only one accession (OsbZIP29 in group A only lost in DG, and OsbZIP59 in group C only lost in R527), indicating that among all the groups containing variable genes, groups A and C were the most stable groups during domestication and improvement of rice. Several OsbZIPs (OsbZIP73, OsbZIP89, OsbZIP38, and OsbZIP93) exist only in one accession, suggesting that these genes may be related to the specific traits of the accessions. Among the 33 accessions, the absence frequency of OsbZIPs in NIP was the lowest (only 4 OsbZIPs). Thirteen genes were lost in G630 and 9311, and the gene absence frequency was the highest in all varieties, which may lead to the differentiation of OsbZIP-related traits between NIP, G630 and 9311.

To explore the selection pressure of OsbZIPs in 33 rice accessions, the Ka/Ks values for the homologous sequence of each OsbZIP in 33 accessions were calculated (Fig. 2). The Ka/Ks values of OsbZIP76 in 02428 and the other 9 accessions are greater than 700; thus, OsbZIP76 was not shown in the figure here, indicating that OsbZIP76 was rapidly evolving due to selection. The Ka/Ks values of most OsbZIPs are between 0 and 1, and the Ka/Ks values of OsbZIP1, OsbZIP37, and OsbZIP85 are far less than 1, indicating that these three genes are stable during evolution and may play a fundamental role in plant growth and development as housekeeping genes; thus, eliminating harmful allelic mutations and under negative purifying selection during evolution. The peaks of Ka/Ks values for OsbZIP10, OsbZIP34, OsbZIP59, OsbZIP62, and OsbZIP84 were located to the right of 1, suggesting that these genes were under positive selection during the domestication of the 33 rice accessions. Out of 94 OsbZIPs, 66 genes do not exhibit Ka/Ks values greater than 1, suggesting that the majority of OsbZIPs underwent purifying selection. The proportion of 28 OsbZIPs with Ka/Ks values greater than 1 in all gene pairs was shown (Fig. 2B). Seven genes (OsbZIP10, OsbZIP14, OsbZIP34, OsbZIP59, OsbZIP62, OsbZIP74, and OsbZIP91) contributes significantly to the proportion of Ka/Ks values greater than 1, suggesting that these genes experienced higher selection pressure during the differentiation process among the 33 accessions and may be undergoing rapid divergence.

Structural Variation is the main source of genetic variation and is closely related to various plant phenotypes [29,30]. Qin et al. [19] identified abundant SVs by aligning 32 high-quality rice genomes with the reference genome (NIP). One hundred and seventy SVs overlap with the 2 kbp downstream, 2 kbp upstream, coding region, and intron regions of the 43 OsbZIPs (Table S2). Aligned with the reference genome, these SVs are classified into deletion, insertion, inversion, and translocation. SVs influence the expression of neighboring genes by either modifying gene sequences or impacting regulatory sequences [31,32]. The effect of SVs on the expression of OsbZIPs was shown based on the RNA-seq data from roots and stems of 33 accessions provided in the study of Qin et al. [19] (Fig. 3, Table S3). Insertion or deletion SVs were located in the intro, 2 kbp upstream, 2 kbp downstream, and coding region of OsbZIP17, OsbZIP18, OsbZIP29, OsbZIP36, OsbZIP57, OsbZIP58, and OsbZIP77 (Fig. 3A), leading to notable alterations in the expression of these genes either in the root or shoot (Fig. 3C). Furthermore, the ratio of SVs that significantly modified the expression levels of OsbZIPs to all SVs overlapping with OsbZIPs was calculated. The results revealed that 4% of SVs overlapping with OsbZIPs significantly altered the expression levels of OsbZIPs (Fig. 3B). These results suggested that the occurrence of some SVs can significantly alter gene expression levels.

To further investigate the effect of SVs on the gene structure of OsbZIPs, TBtools was used to analyze the gene structure of OsbZIPs whose expression levels were significantly altered by SVs (Fig. 4 and Figure S1). The gene structures of SV-influenced OsbZIPs were different in different varieties. Two genes (OsbZIP29 in 9311 and OsbZIP36 in Tumba) do not contain any motif, and thus these two genes cannot be identified in the reference genome-based gene family identification and would be neglected in functional studies. This problem may be avoided by identifying gene families with the gene-based pan-genome and making the gene family members becoming more comprehensive by colinear relation among 33 rice accessions (Table S4). It could provide powerful information for assigning functions to OsbZIPs. Moreover, due to the occurrence of SVs, some genes, such as OsbZIP29 in D62 and R498, were undergoing gene truncations. This may lead to interesting changes in gene functions, such as pseudogenization or the emergence of new functions [33], which deserve further study.

To investigate how SVs impact the conserved domains of OsbZIPs across different varieties, the motifs of OsbZIPs in FS32, which has the highest number of SV overlapped OsbZIPs, were displayed (Fig. 5A). Four of the 10 motifs of OsbZIPs in FS32 did not correspond with the reference. The amino acids in each set of mutually corresponding motifs, as well as the consistency of each mutually corresponding amino acid are not the same. These results indicate that the motifs of OsbZIPs were strongly influenced by SVs. As gene family identification relies on conserved domain analysis, alterations in these domains often lead to the neglect of these genes. Hence, we counted typical genes (containing conserved domains) and atypical genes (lacking conserved domains) in each accession (Fig. 5B). Six OsbZIPs (OsbZIP25, OsbZIP31, OsbZIP48, OsbZIP81, OsbZIP88, and OsbZIP90) had both typical and atypical genes in most varieties; while OsbZIP52 was present as atypical genes in 32 varieties except N22 which had both typical and atypical genes. These results suggest that SVs affect the conserved domain of OsbZIPs and lead to the emergence of a large number of atypical genes.

Rice bacterial blight (BB) is caused by Xoo infection, which leads to a reduction in rice yield [34]. Increased temperature due to global warming suppresses host plant defense against the pathogen. Dossa et al. [35] infected a susceptible line (IR24) and a resistant line (IRBB67) with Xoo under high temperature and performed RNA-seq at 3, 72 and 120 hpi. To explore the response of OsbZIPs to BB, the gene expression levels of IR24 and IRB667 at different times of mock and Xoo inoculation were compared and there were 34 OsbZIPs genes were differentially expressed (Fig. 6, Table S5). Compared to the genes with differential expression in Xoo-infected and mock, there were more genes (44) with differential expression at different time points post-infection than Xoo-infected or mock (25), indicating a more pronounced response of OsbZIP to Xoo infection at different infection times (Figs. 6B–6F). In the comparison between Xoo-infected and mock groups across various temperatures and samples at 3 hpi (Figs. 6C–6F), a higher number of differentially expressed OsbZIPs were observed, suggesting a critical response period for certain OsbZIPs to Xoo at 3 hpi. Moreover, Only OsbZIP75 showed differential expression among different varieties, indicating that OsbZIPs’ response to Xoo was minimally influenced by sample variations. Only 3 OsbZIPs exhibited differential expression at 120 hpi in IR24 under high vs. low temperature conditions, suggesting a limited impact of temperature variations on OsbZIPs’ response to Xoo infection in IR24.