Background

Throughout evolution, organisms have developed intrinsic biological clock systems to effectively adjust and acclimate to the cyclic alterations in Earth’s environment [1,2,3,4]. The core components of the biological clock rhythmically regulate downstream gene expression through transcription-translation negative feedback loops (TTFL) to maintain circadian rhythms of life activities [5,6,7,8]. The core biological clock system regulates the circadian rhythm of histone modifications in the genome by controlling the diurnal oscillations of gene expression responsible for initiating histone modifications [9, 10]. Active histone modifications enhance the binding activity of cis-regulatory transcription factors to the genome, facilitating the recruitment of RNAPII to the transcription start sites. This promotes the transcription activation of rhythm genes throughout the genome, thereby maintaining the diurnal oscillation of genome-wide transcription.

Numerous studies have combined research investigations on the circadian rhythm of animals, plants, and microorganisms with high-throughput technologies to explore various perspectives, including transcriptome, histone modification, DNA methylation, transcription factor recruitment, proteome, and metabolism [4]. In mouse liver circadian rhythm studies, mRNA sequencing (RNA-seq) and chromatin immunoprecipitation sequencing (ChIP-seq) identified genome-wide cis-regulatory elements of the core biological clock components [7]. These studies confirmed that histone modifications’ diurnal oscillations can promote the gene expression’s diurnal oscillation [11]. Furthermore, the analysis of RNAPII recruitment and rhythmic gene expression revealed dynamic changes in gene expression during the circadian cycle, and the related epigenetic modification of circadian remodeling [12]. Diurnal RNA-seq experiments in multiple organs revealed circadian transcription heterogeneity across tissues and gene diurnal oscillations’ generality [13]. In B. rapa, gene coexpression network analysis of rhythmic genes found that the circadian fluctuation of gene expression is extensively altered under drought stress, thus identifying drought-responsive genes [14]. RNA-seq of diurnal oscillations in soybean revealed that the biological clock responds to abiotic stress by regulating the phase of gene oscillations [15]. Circadian oscillating RNA-seq in Arabidopsis shows a circadian rhythm in clock-regulated alternative splicing [16]. In Arabidopsis, studies of the dynamics of histone modifications have described that H3K27me3 and H3K4me3 exhibit distinct plasticity and stability in diurnal and seasonal fluctuations, respectively [17]. The circadian changes in protein-coding gene expression in plants depend on the dynamic reversible acetylation and phosphorylation of histone modifications in corresponding genomic regions, and reduced binding of repressive transcription factors contributes to activating-property chromatin modifications [18]. Recently, developed technologies such as Chromatin Interaction Analysis with Paired-End Tag (ChIA-PET) and high-throughput chromosome conformation capture (Hi-C) have brought circadian rhythm research into the perspective of three-dimensional space. The circadian rhythm of the three-dimensional genome structure dynamically configures the regulatory elements of gene expression [19,20,21]. In rice, the circadian dynamics of RNAPII-mediated chromatin interactions correlate with diurnal oscillations in gene expression [22]. Unlike mammals, plants lack the mammalian-like suprachiasmatic nucleus (SCN) of the hypothalamus in which the core circadian oscillator functions [2]. The unique fixed growth pattern and seasonal growth and development characteristics of plants mean that the mechanism by which plants sense periodic changes in the external environment is more sensitive and complex. Similar to the mammalian peripheral clock, the circadian rhythm characteristics of physiological activities in different plant organs are also tissue-specific and related to different light inputs [23]. Chromatin accessibility likewise differs in different Arabidopsis tissues in response to photoperiods [24]. The expression of biological clock genes exhibits tissue specificity with asymmetric coupling [25]. Dispersed plant biological clocks process environmental signals separately in specific tissues, and there is a clear division of labor between biological clocks in each tissue, which coordinates the overall circadian physiological activities of the plant [26]. However, the characterization of diurnal oscillations in epigenetic modifications is still lacking in plant circadian rhythm studies.

Polyploid species typically possess multiple copies of biological clock genes, and elucidating the homologous expression diversity of rhythmic genes can aid in unraveling the complex biological functions in which biological clock genes are involved. The alignment of known biological clock genes between Arabidopsis thaliana and B. rapa has revealed that biological clock genes are preferentially retained during gene loss after multiple rounds of genome doubling events [27]. Studies conducted on three different plant morphological Brassica species have demonstrated that numerous copies of biological clock genes are selectively retained after a genome doubling event to diversify their functions, thereby contributing to the species diversity of Brassica [28]. Atlantic salmon has undergone 4 rounds of whole-genome duplication since vertebrates, and the diversification and tissue-specific expression of multicopy biological clock genes are associated with functions other than the biological clock, such as gill development processes that adapt to migration [29]. Approximately 7500 years ago, the A genome of Brassica rapa (B. rapa, 2n=2x=20, AA) and the C genome of Brassica oleracea (B. oleracea, 2n=2x=18, CC) hybridized naturally and formed Brassica napus (B. napus, 2n=4x=38, AACC) through genome doubling [30, 31]. Due to the imbalance of asymmetric transcription and epigenetic modification, gene expression activity in the An subgenome was significantly higher than that in the Cn subgenome in B. napus [32]. The genetic background of allopolyploidy and the recent origin of B. napus make it an excellent model for exploring differences in diurnal oscillations between subgenomes and homologous genes.

To characterize the diurnal oscillation characteristics of the B. napus genome, we generated high-quality gene expression and histone modification datasets using RNA-seq and ChIP-seq before the onset of flowering. We discovered that the expression of protein-coding genes, histone modifications (H3K4me3, H3K9ac), and RNAPII recruitment exhibited robust diurnal oscillations. While the gene expression and active histone modifications of the An subgenome were higher than those of the Cn subgenome [32], we observed that the diurnal oscillation of the Cn subgenome was more potent than that of the An subgenome. Compared to B. rapa, one of B. napus’ ancestors, we found extensive changes in the A genome’s circadian oscillatory signature of genes. Furthermore, we found that the differences in diurnal oscillations of expression of homologous genes were associated with changes in various histone modifications during the circadian cycle. Our study analyzes the circadian oscillatory signature of transcription and histone modifications in allopolyploid species, providing insights into genome duplication and the resulting redundancy of biological clock genes, diurnal oscillation differences between subgenomes, and differences in the diurnal oscillation of expression of homologous genes.

Results

Robust diurnal oscillation of protein-coding genes in B. napus

We collected samples every 4 h over 2 full diurnal cycles under stable meteorological conditions (Additional file 1: Fig. S1) to generate high-quality transcriptomes of B. napus by RNA-seq. We aimed to detect the circadian oscillatory signature of protein-coding gene expression. Using one of the latest high-quality reference genome assembly results [33], we mapped 96,921 protein-coding genes in allotetraploid B. napus on 10 chromosomes in the An subgenome and 9 chromosomes in the Cn subgenome. To quantify gene expression, we used a deep neural network (DNN)-based machine learning method called BIO_CYCLE [34, 35] and measured transcripts per million (TPM). We detected 22,445 rhythmic genes (23.16%, out of 96,921 protein-coding genes) with q values <0.05 and a period of 20 to 28 h (Fig. 1C). Among the protein-coding genes, 18,623 genes (19.21%) were not expressed within 1 week before the flowering transition (Fig. 1A). We also detected 63,016 rhythmic exons (14.00%, of 450,037 exons) and 5491 rhythmic introns (4.96%, of 110,639 introns) at the transcript level (Fig. 1G). We also investigated diurnal oscillations at the transcript level. Throughout the genome, 22,061 genes exhibited diurnal oscillations in their exonic expression, with 97.17% (21,436) classified as rhythmic genes. Additionally, 3817 genes showed diurnal oscillations in their intronic expression, with 86.61% (3,306) classified as rhythmic genes. A total of 3143 genes exhibited diurnal oscillations in both exonic and intronic regions. However, due to variations in the oscillation phases of complex transcripts, 99.52% (3,128) of these are categorized as rhythmic genes (Fig. 1B). Since intronic signals can be interpreted as pre-mRNA expression or nascent transcriptional events [36], and exonic signals as mRNA transcriptional events [7], we believe that the diurnal oscillation of gene expression can be attributed to the diurnal oscillation of exons (Fig. 1E).

Fig. 1
figure 1

Diurnal oscillation characteristics of protein-coding gene transcription in B. napus. A Proportions of genes with different expression signatures of protein-coding genes in B. napus. B The number of genes with overlap** rhythmic exons and introns. C The heatmap illustrates the diurnal oscillation of gene expression in rhythmic genes. D A frequency plot percentage of rhythmic genes with peak expression phase binned in 4h over 24h is presented. E The phase distribution of rhythmic genes, rhythmic exons, and rhythmic introns. F Spatial-specific clustering of time-series curves of rhythmic gene transcription. G Overlap** genes with oscillating exons and introns overlap with rhythmic genes. H Features of gene transcription at the indicated time point of LHY. (From left to right, they are BnaA10G0008600ZS, BnaC03G0001300ZS, and BnaC05G0010100ZS) I Diurnal fluctuations in high-amplitude homologous gene expression of LHY. (The red line represents BnaA10G0008600ZS, the blue line represents BnaC03G0001300ZS, and the green line represents BnaC05G0010100ZS) J Comparing amplitude, peak-to-trough ratio, and median of rhythmic gene expression with global gene expression

Full size image

We partitioned 24 h into 6 intervals of 4 h and calculated the phase distribution of all rhythmic gene expression peaks. Among them, 45.59% of rhythmic gene expression peaks occurred during the daytime (8:00–16:00), and the ratios of genes expressed during dusk (16:00–20:00) and early morning (4:00–8:00) were similar. Similarly, the ratios of genes expressed during the morning (8:00–12:00) and afternoon (12:00–16:00) were similar, as were the ratios of genes expressed during eve (20:00–24:00) and dawn (0:00–4:00) (Fig. 1D). The median, amplitude, and fold change of rhythmic genes were higher than those of globally expressed genes (Fig. 1J). Gene Ontology (GO) analysis revealed that rhythmic genes had extensive binding and catalytic activities and were involved in biological processes such as metabolism and cellular processes (Additional file 1: Fig. S3C). Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis indicated that most rhythmic genes had kinase activity (Additional file 1: Fig. S3D). Clustering analysis of rhythmic genes based on time-series spatial similarity features fully revealed differences in gene diurnal oscillation features (Fig. 1F).

As an allopolyploid species, B. napus typically has multiple copies of genes. LHY encodes a critical transcription factor involved in the plant’s biological clock and can bind to the promoter regions of TOC1 and CHE to inhibit their transcription, thereby suppressing the expression of CCA1. In B. napus, 8 copies of LHY are distributed on 6 chromosomes in the An/Cn subgenomes. Among the 5 rhythmic copies of LHY, 3 are categorized as high-amplitude rhythmic genes (Fig. 1H, I), while the remaining 2 display lower oscillation amplitudes (Additional file 1: Fig. S3B).

Epigenetic modifications with activating properties promote diurnal oscillations in gene expression

We produced a collection of circadian rhythmic histone modifications data, encompassing H3K4me3 and H3K9ac, both linked to the activation of gene expression, along with RNAPII recruitment, a participant in gene expression, within B. napus under identical conditions to those of the diurnally fluctuating transcriptome dataset. We observed robust diurnal oscillation characteristics for histone modifications using the same diurnal oscillation detection method as the transcriptome dataset (Fig 2I–K). We compared histone modifications’ dynamic abundance at expressed, non-rhythmic, and non-expressed genes (Additional file 1: Fig. S6C-E; S7C-E; S8C-E). Furthermore, we compared the differences in histone modifications between rhythmic and non-rhythmic genes at the same time, as well as between expressed genes and non-expressed loci (Additional file 1: Fig. S6G, H; S7G, H; S8G, H). Although histone modifications are related to gene expressional activation, their phase distributions were significantly different (Fig. 2E). Moreover, the clustering analysis of diurnal oscillation variations revealed distinctions in temporal sequence features and oscillatory characteristics (Additional file 1: Fig. S5A-C).

Fig. 2
figure 2

Diurnal oscillation patterns of epigenetic modifications. A Ratio of rhythmic peaks in histone modifications H3K4me3, H3K9ac, and RNAPII recruitment. B The rhythmic peaks of histone modifications H3K4me3, H3K9ac, and RNAPII recruitment are distributed in different rhythmic feature genes. C Annotation of rhythmic peaks for histone modifications H3K4me3, H3K9ac, and RNAPII recruitment. D Overlap** genes with rhythmic peaks in histone modifications H3K4me3, H3K9ac, RNAPII recruitment, and rhythmic genes. E Phase distribution of rhythmic peaks in histone modifications H3K4me3, H3K9ac, and RNAPII recruitment. F Phase differences between rhythmic peaks and overlap** genes of histone modifications H3K4me3, H3K9ac, and RNAPII recruitment. G Heatmap of rhythmic peaks modified by histone modification H3K9ac and their gene transcription. H Diurnal oscillation of COL9 homologous gene expression and its histone modifications H3K4me3, H3K9ac, and RNAPII recruitment. I Diurnal oscillation of histone modification H3K4me3 on rhythmic genes. J Diurnal oscillation of histone modification H3K9ac on rhythmic genes. K Diurnal oscillations of RNAPII recruitment on rhythmic genes

Full size image

We identified 53,738 H3K4me3 peaks throughout the genome, of which 13.55% (7284) were diurnal oscillations (Fig. 2A, B). Peak annotation results revealed that 1268 rhythmic H3K4me3 peaks are located within the promoter regions of protein-coding genes (Fig. 2C). There are 622 H3K4me3 peaks located within the promoters of 622 rhythmic genes (Fig. 2B, Additional file 1: Fig. S4A). GO enrichment analysis indicated these genes were mainly associated with metabolic and catalytic activity (Additional file 1: Fig. S4G). For H3K9ac, we detected 78,069 peaks throughout the genome, of which 14.23% (11,112) were diurnal oscillations (Fig 2A, B). Peak annotation results reveal that 2153 rhythmic H3K9ac peaks are within the protein-coding gene promoter regions (Fig. 2B, C). There are 1145 H3K4me3 peaks located within the promoters of 1141 rhythmic genes (Fig. 2B, Additional file 1: Fig. S4B). GO enrichment analysis revealed these genes were broadly binding, catalytically active, and involved in biological processes such as metabolism and cells (Fig. S4I). Activating histone modifications increases RNAPII recruitment to promote gene expression [37]. We identified 71,166 peaks throughout the genome, of which 15.00% (10,678) were diurnal oscillations (Fig 2A, B). Annotation results indicate the presence of 9544 rhythmic RNAPII peaks across the gene bodies of 9265 protein-coding genes (Fig. 2B). Among them, 5600 genes, accounting for 60.44%, were identified as rhythmic genes (Additional file 1: Fig. S4C). GO enrichment analysis revealed that these genes are involved in metabolic and cellular processes, with catalytic and transcription factor binding activities (Additional file 1: Fig. S4K). The peak phase distributions of the histone modifications H3K4me and H3K9ac on diurnal oscillatory peaks were similar, with a higher ratio during the day (8:00–16:00) than at night (20:00–4:00) (Additional file 1: Fig. S4D, E). However, the peak phase of histone modifications by RNAPII recruitment was mainly concentrated at dusk (16:00–20:00) and the eve of the night (20:00–24:00) (Additional file 1: Fig. S4F). The diurnal oscillation of histone modification H3K9ac and RNAPII recruitment preceded that of protein-coding genes (Fig. 2F, G; Additional file 1: Fig. S4J, L; S11D, E). Although associated with the transcriptional activation of genes, the diurnal oscillation of histone modification H3K4me3 peaks occurred later than that of gene expression (Fig. 2F; Additional file 1: Fig. S4H; S11B, C).

The histone modifications H3K4me3 and H3K9ac are associated with the activation of gene expression, while RNAPII directly participates in this process. Among rhythmic genes, 29.41% (6602) exhibited at least 1 type of histone modification with diurnal oscillations (Fig. 2D). As previously reported [11], H3K4me3 and H3K9ac, which facilitate diurnal oscillations in transcription, exhibit robust diurnal oscillations not only in promoter regions, but also in other elements of the genome (Fig. 2C). However, some genes with diurnal oscillation modifications do not express, manifesting as an accumulated expression level of 0 during the day-night cycle (Fig. 2B). Diurnal oscillations of histone modifications are not exclusive to rhythmic genes, but are widespread across the genome. We identified histone modification peaks in a total of 55,038 genes, representing 56.78% of all protein-coding genes. Among these genes, 16,915 have at least 1 type of histone modification peak that oscillates day and night, and the expression of 9351 genes also oscillates. Although H3K4me3 activates promoters, H3K9ac activates promoters and enhancers, and RNAPII is directly involved in transcription, our results demonstrate that the diurnal oscillating histone modifications detected in different regions of protein-coding gene bodies contribute to varying degrees of diurnal oscillations in transcription.

We examined the correlation between histone modification elements and gene expression using Pearson’s correlation coefficient (R). The diurnal oscillation of H3K4me3 modification peaks showed a positive correlation with gene expression diurnal oscillation (R=0.54). The correlation coefficient of promoter and exon regions was slightly lower than the global average, while intron regions showed the highest correlation (R=0.64), and transcription termination regions were the least correlated (R=0.31) (Additional file 2: Supplementary Dataset 1). The diurnal oscillation of H3K9ac modification peaks was also positively correlated with gene expression diurnal oscillation (R=0.58), with the correlation coefficient at the promoter region slightly higher than that at the global level (R=0.59). Exon regions showed the highest correlation (R=0.61), while intron regions had a moderate correlation with transcription termination regions (0.45<R<0.46) (Additional file 3: Supplementary Dataset 2). The diurnal oscillation of RNAPII modification peaks was also positively correlated with gene expression diurnal oscillation (R=0.51). The correlation coefficients of the promoter region, exon region, intron region, and transcription termination region were all close (0.43<R<0.52) (Additional file 4: Supplementary Dataset 3). The diurnal oscillations of epigenetic modifications associated with gene expression activation exhibit varying degrees of correlation with the diurnal oscillations of protein-coding gene expression. Moreover, these correlations display distinctions among different elements. Furthermore, differences in chromatin accessibility were observed between rhythmic and non-rhythmic genes, as well as expressed and non-expressed genes (Additional file 1: Fig. S9A, C). Motif analysis of rhythmic, non-rhythmic, and non-expressed genes revealed differences between discrete gene classes (Additional file 1: Fig. S9B, D, and F).

Using H3K4me3 as a defining mark, we performed a preliminary study to explore the diurnal oscillations of epigenetic modifications on core plant clock genes in B. napus, including COL2, COL9, GI, LNK1, PIF4, PRR3, PRR7, RVE4, SPA1, and TIC. H3K9ac is also involved in modifying the promoter regions of core clock genes, including CBF2, CCR1, CDF2, COL9, COR27, COR28, GI, HYH, LNK1, LNK2, PHYB, PIF4, PRR3, PRR5, PRR7, PRR9, RVE1, RVE2, RVE3, RVE4, and TIC (Additional file 5: Supplementary Dataset 4). COL9, which belongs to the CO gene family and regulates the expression of FT and SOC1, has 5 homologous copies. We observed that the promoter region of BnaA05G0456200ZS exhibits diurnal oscillations in both H3K4me3 peaks (R=0.42) and H3K9ac peaks (R=0.80) (Fig. 2H).

Diurnal oscillation of the Cn subgenome exhibits a stronger amplitude than that of the An subgenome

We compared the oscillatory properties of rhythmic genes between the An/Cn subgenomes of B. napus. Our findings suggest that the rhythmic genes on the An/Cn subgenomes had similar distribution densities in phase shift, amplitude, and period (Fig. 3A–C). While there was a significant difference in the distribution of expressed genes between the An/Cn subgenomes, there was no difference in the distribution of rhythmic genes (Additional file 1: Fig. S9G). Integrating the epigenetic modification dataset, we observed differential diurnal oscillation strengths in active epigenetic modifications between the An/Cn subgenomes. Moreover, the diurnal oscillation peaks of epigenetic modifications in the Cn subgenome were found to be higher than those in the An subgenome (Additional file 1: Fig. S6A, B, and F; S7A, B, and F; S8A, B, and F). In contrast to the Cn subgenome, the An subgenomes exhibit higher chromatin accessibility at the transcription start sites and upstream regions of rhythmic genes. The presence of large and small peaks suggests the potential existence of multiple regulatory elements, such as enhancers or transcription factor binding sites (Additional file 1: Fig. S9E).

Fig. 3
figure 3

Diurnal oscillation disparities between subgenomes and the evolution of their characteristics. A–C Density distributions of phase shifts, amplitudes, and periods of rhythmic genes in the An/Cn subgenome of B. napus. D–F Density distributions of phase shifts, amplitudes, and periods of rhythmic genes in the B. rapa A genome and the B. napus An subgenome are presented. G The proportion of genes with different expression characteristics between homologous gene pairs and subgenome-specific genes. (Red represents non-rhythmic genes, blue represents rhythmic genes, and green represents unexpressed genes) H Chromatin accessibility disparities between homologous genes and subgenome-specific genes. I The expression heatmap of rhythmic genes in the B. rapa A genome and homologous genes in the B. napus An subgenome. J Differences in histone modifications between homologous genes and subgenome-specific genes

Full size image

We compared the differences in rhythmic genes between two Brassica species, B. napus and B. rapa, explicitly focusing on the A genome. We used the published diurnal oscillation dataset from the natural environment and applied the same diurnal oscillation detection method as in B. napus [14, 38]. The An subgenome of B. napus contains 44,359 genes, of which 36,389 are expressed, and 10,619 are rhythmic. The A genome of B. rapa has 45,409 genes, of which 36,344 are expressed, and 15,314 are rhythmic. We also compared the changes in the An subgenome’s rhythmic gene day-night oscillation characteristics during the evolution from B. rapa to B. napus. The phase shift of rhythmic genes showed extensive variation (Fig. 3D), indicating that after encountering the A/C genomes, functional gene redundancy and chromosomal variations affected the temporal expression patterns of most genes. While the density distribution curves of the amplitudes and periods of rhythmic genes in B. rapa and B. napus showed certain similarities, the distantly hybridized and genome-duplicated events led to a noticeable trend of higher numbers of low-amplitude genes and reduced numbers of high-amplitude genes (Fig. 3E). The oscillation periods of rhythmic genes also exhibited a trend of differentiation, either shorter or longer than 24 h (Fig. 3F). We aligned the protein sequences of rhythmic genes in B. rapa with their homologous genes in the An subgenome of B. napus. The diurnal oscillatory heatmap of gene expression showed phase advancement and widespread diurnal oscillatory pattern changes (Fig. 3I). The proportion of genes expressed in An subgenome of B. napus during the day-night cycle was significantly higher than that of B. rapa. In comparison, the proportion of rhythmic genes was substantially lower than that of the A genome of B. rapa (Additional file 1: Fig. S9H).

The encounter of the A genome of B. rapa and the C genome of B. oleracea, prior to subsequent polyploidization in B. napus, resulted in the redundancy of functional genes. The gene dosage hypothesis suggests that the functions of homologous genes in B. napus would produce a trend of diversification [39]. To investigate this, we constructed a list of homologous genes across 4 species: Arabidopsis thaliana, B. rapa, B. oleracea, and two subgenomes of B. napus, based on the protein sequences of Arabidopsis protein-coding genes. The protein-coding genes of B. napus were divided into subgenome-homologous gene pairs and subgenome-specific genes. We compared the diurnal oscillations of these two groups and found that the proportion of subgenome-homologous gene pairs in the middle rhythmic gene was as high as 40%. In contrast, the proportion of rhythmic genes in the subgenome-specific genes was only 3%, with the latter having a large proportion of genes that were not expressed (Fig. 3G). Differences in chromatin accessibility of homologous genes versus subgenome-specific genes were observed in the ATAC-seq data from the seedling stage of B. napus (Fig. 3H). Furthermore, combined with ChIP-seq, we found that the promoter activation-associated histone modifications H3K4me3 and H3K9ac, as well as the RNAPII recruitment involved in transcription, had higher modification peaks on homologous gene pairs (Fig. 3J; Additional file 1: Fig. S11A).

Alteration of diurnal oscillatory characteristics of biological clock homologous genes during allopolyploidization

We searched for 77 genes directly involved in the core biological clock regulatory network of Arabidopsis thaliana. Homologous alignment of protein sequences was used to blast for homologous copies of biological clock genes in B. rapa, B. oleracea, and B. napus. There are 146 homologous circadian genes in the B. rapa genome, 170 biological clock homologous genes in the B. oleracea genome, and 312 biological clock homologous genes in the B. napus genome. The allopolyploid species B. napus inherited all homologous circadian genes from B. rapa and B. napus (Additional file 5: Supplementary Dataset 4). However, after distant hybridization and polyploidization, the copy number of core biological clock homologous genes in B. napus varied. For instance, the biological clock genes ELF4 and CDF6 have several homologous copies in B. rapa and B. oleracea but are entirely lost in B. napus. Compared to B. rapa and B. oleracea, the copy numbers of some biological clock homologous genes in B. napus have changed due to duplication and deletion. LHY has 3 consecutive homologous copies on chromosome C7 and one copy on chromosome C8 in B. rapa, whereas in B. napus, 1 copy was missing on chromosome C7, and one copy was added adjacent to the homologous gene on chromosome C8. The variation in chromatin structure in B. napus has also led to the translocation of biological clock homologous genes between chromosomes. LWD1 has a copy on chromosome C5 of B. oleracea, but in B. napus, it produces two copies located on chromosomes C6 and C7. Interestingly, the biological clock genes ZTL, FKF1, and LUX-like were deleted in all three Brassica species. In summary, the majority of biological clock homologous genes in B. napus are inherited conservatively, both in terms of chromosomal location and quantity, from their counterparts in B. rapa and B. napus.

In our circadian transcription dataset, 173 out of 312 biological clock homologous genes in B. napus exhibit diurnal oscillation properties. Among these genes, 17 homologous copies of biological clock genes lose their circadian rhythm entirely in B. napus, while another 60 biological clock genes maintain their transcriptional circadian rhythms, with at least 1 copy of these genes exhibiting diurnal oscillations. To further investigate the diurnal oscillation characteristics of biological clock genes in B. rapa and B. napus, we compared the expression patterns of homologous copies of these genes in the 2 species using published circadian rhythm transcriptome datasets. Our findings reveal that homologous copies of biological clock genes, demonstrating diurnal oscillations, frequently exhibit distinct amplitudes while maintaining similar phases (Additional file 1: Fig. S10).

Differential epigenetic modifications contribute to variations in the rhythmic expression of homologous genes

Among the copies of homologous genes in the biological clock of B. napus, there were differences in the characteristics of diurnal oscillations, such as amplitude, phase, and expression level. Additionally, we observed differences in the histone modification patterns among the copies of clock genes.

The expression of all 5 homologous genes of COL9 in B. napus exhibits diurnal oscillations. Diurnal oscillation modification signals for BnaA05G0456200ZS and BnaC03G0371800ZS are located in the promoter region, and the transcriptional diurnal oscillations of these 2 homologous genes differ (Fig. 4H, I). This difference is associated with H3K4me3 (Fig. 4B, C). PIF4 is a negative regulator of the plant red light response and participates in plant shade avoidance. There are 4 homologous copies of PIF4 in B. napus, and the expression of 3 copies oscillates diurnally. We detected diurnal oscillations of H3K4me3 modification in the promoter regions of BnaC03G0245900ZS and BnaC04G0590800ZS (Fig. 4D, E), which maintained a similar phase to the diurnal oscillations of gene expression but had differences in amplitude (Fig. 4J, K). PRR7 is a functionally redundant component of the temperature-sensitive circadian system. CCA1 and LHY positively regulate its expression, and it acts as a transcriptional repressor of both genes, regulating hypocotyl elongation through the photoperiodic pathway. There are 5 homologous copies of PRR7 in B. napus, and the expression of 4 genes oscillates around the clock. We detected diurnal oscillations of H3K4me3 modification in the promoter regions of BnaA02G0009400ZS and BnaC09G0614800ZS (Fig. 4F, G). The histone modification signal of BnaA02G0009400ZS was approximately twice that of BnaC09G0614800ZS.

Fig. 4
figure 4

Variation in rhythmic expression of biological clock homologous genes is associated with differences in diurnal oscillations of histone modification H3K4me3. A Features of diurnal oscillation of histone modification H3K4me3 on COL9 (BnaA05G0456200ZS, BnaC03G0371800ZS), PIF4 (BnaC03G0245900ZS, BnaC04G0590800ZS), and PRR7 (BnaA02G0009400ZS, BnaC09G0614800ZS). B–G Diurnal oscillation of histone modification H3K4me3 on COL9 (BnaA05G0456200ZS, BnaC03G0371800ZS), PIF4 (BnaC03G0245900ZS, BnaC04G0590800ZS), and PRR7 (BnaA02G0009400ZS, BnaC09G0614800ZS). H–M Diurnal fluctuations in homologous gene expression of COL9 (BnaA05G0456200ZS, BnaC03G0371800ZS), PIF4 (BnaC03G0245900ZS, BnaC04G0590800ZS), and PRR7 (BnaA02G0009400ZS, BnaC09G0614800ZS)

Full size image

HYH is involved in the red light response of plant photosynthesis. All 6 homologous genes of HYH in B. napus exhibit diurnal oscillation, and histone modification H3K9ac signals that oscillate day and night were detected. The modification signals of BnaA01G0333200ZS, BnaA05G0383200ZS, and BnaC01G0412000ZS are located in the promoter region (Additional file 1: Fig. S12A). The diurnal oscillations of gene transcription are highly correlated with the diurnal oscillations of histone modification signals (Additional file 1: Fig. S12B-D, I-K). The transcription factor encoded by RVE1 regulates hypocotyl elongation by controlling the auxin level during a specific period. The expression of all 6 homologous genes of this gene in B. napus also exhibits diurnal oscillation. The promoter regions of BnaA10G0195900ZS and BnaC02G0079400ZS exhibit diurnal oscillations of histone modification H3K9ac signals, and the diurnal oscillations of gene expression are highly correlated with the diurnal oscillations of histone modification signals (Additional file 1: Fig. S12E, F, L, and M). RVE3 is involved in the blue light response, and 3 homologous genes of RVE3 are found in B. napus, namely BnaA10G0004600ZS, BnaA10G0004700ZS, and BnaC05G0006800ZS, all of which exhibit diurnal oscillation and function as circadian rhythm genes. However, it should be noted that BnaA10G0004600ZS shows relatively low cumulative expression levels. There was a diurnal oscillation of histone modification H3K9ac signals in the promoter region of the gene, and the diurnal fluctuation of gene expression was highly correlated with the diurnal oscillation of histone modification signals (Additional file 1: Fig. S12G, H, N, and O).

CHE has 6 homologous genes in B. napus, 5 exhibiting diurnal oscillation in transcription. We detected oscillatory recruitment signals of RNAPII at the transcription start sites of 3 of these genes (Additional file 1: Fig. S13B, C, H, and I). LHY has 8 homologous genes in B. napus, 5 exhibiting diurnal oscillation in transcription. We also detected oscillating RNAPII recruitment signals at the transcription start sites of 2 of these genes (Additional file 1: Fig. S13D, E, J, and K). Similarly, all 6 homologous genes of RVE1 in B. napus exhibit diurnal oscillation in transcription, and we detected oscillatory RNAPII recruitment signals at the transcription start sites of 3 of these genes (Additional file 1: Fig. S13F, G, L, and M). Although RNAPII is involved in gene expression, we did not detect RNAPII recruitment signals at any diurnally oscillating gene loci. However, for a small portion of diurnally oscillating genes, there was a high correlation between the diurnal oscillations and RNAPII recruitment signals.

Diverse epigenetic modifications define diurnal oscillations of gene expression

The genome harbors a wide array of histone modifications, with various combinations often observed at a specific locus. The diurnal oscillation of rhythmic gene expression is occasionally highly correlated with more than one oscillation of histone modifications. We integrated a table of histone modification data for all protein-coding genes (Additional file 6: Supplementary Dataset 5). When multiple diurnal oscillation histone modifications were identified on rhythmic genes, different types of histone modifications typically exhibited similar diurnal oscillation phases. In conclusion, numerous histone modifications may simultaneously affect the transcription of rhythmic genes. LNK1 plays a role in integrating plant light signaling and the biological clock, which is regulated by the TOC1 complex and acts as a transcriptional coactivator. In B. napus, there are 4 homologous copies of LNK1, of which BnaC02G0541300ZS is a non-expressed gene (Fig. 5E), and no peaks of histone modifications were detected for this gene (Fig. 5F–H). BnaA06G0292400ZS, BnaC03G0534100ZS, and BnaC09G0081200ZS are rhythmic genes (Fig. 5E). Among them, we observed diurnal oscillations of histone modification H3K9ac and RNAPII recruitment signals on BnaA06G0292400ZS (Fig. 5G, H), while histone modification H3K4me3 signals displayed diurnal oscillations (Fig. 5F). The expression of this gene exhibits a higher correlation with the recruitment signals of RNAPII during day-night oscillations. The histone modifications H3K4me3 and H3K9ac on BnaC03G0534100ZS both undergo day-night oscillations. However, the diurnal gene expression variation correlates more with the day-night oscillations of histone modification H3K9ac (Fig. 5F–H). Despite the day-night oscillations of histone H3K4me3 and H3K9ac modifications on BnaC09G0081200ZS, the oscillations of H3K4me3 are out of phase with gene expression (Fig. 5F). At the same time, while there is a higher correlation between the day-night oscillations of H3K9ac and transcription (Fig. 5G), while RNAPII signals were not detected in BnaC09G0081200ZS (Fig. 5H).

Fig. 5
figure 5

Combinations of multiple epigenetic modification define the diurnal oscillations of gene expression. A Features of diurnal gene expression oscillation in 4 homologous genes of LNK1. B Features of diurnal histone modification H3K4me3 oscillation in 4 homologous genes of LNK1. C Features of diurnal histone modification H3K9ac oscillation in 4 homologous genes of LNK1. D Features of diurnal RNAPII recruitment oscillation in 4 homologous genes of LNK1. E Diurnal oscillation of expression in 4 homologous genes of LNK1. F Diurnal oscillation of histone modification H3K4me3 in 4 homologous genes of LNK1. G Diurnal oscillation of histone modification H3K9ac in 4 homologous genes of LNK1. H Diurnal oscillation of RNAPII recruitment in 4 homologous genes of LNK1

Full size image

We observed a complex combination of histone modifications on RVE1. Diurnal oscillations of both H3K4me3 and H3K9ac modifications were detected on all 6 homologous genes of RVE1 in B. napus (Additional file 1: Fig. S14). The diurnal oscillations in the expression of all 6 homologous genes were highly correlated with the diurnal oscillations of H3K9ac modification. We observed that the peak of histone modification H3K4me3 occurs later than the peak of gene expression. In contrast, the peak of histone modification H3K9ac occurs earlier than the peak of gene expression, which is consistent with the results from the genomics analysis section (Fig. 2G). Only 4 copies showed circadian signals of RNAPII recruitment, which were also highly correlated with the diurnal oscillation of gene expression.

Discussion

Our study presents a high-resolution, multiperiod time-series dataset from B. napus that characterizes diurnal oscillations of multiple histone modifications associated with global transcription and gene expression activation in allopolyploid plants. Diurnal oscillations of gene expression play a crucial role in enabling plants to regulate their physiological activities and successfully adapt to the cyclical changes in the Earth’s environmental circadian cycle. Diurnal oscillations of histone modifications contribute to the diurnal oscillation of genes at the transcriptional level, and the superposition of multiple histone modifications jointly enhances the robustness of diurnal oscillations of gene expression. This mechanism enables plants to achieve superior stability in regulating their life activities and successfully adapt to the changes in the external circadian environment [1, 40, 41]. During the process of interspecific hybridization and allopolyploidy, the global circadian rhythm of the genome undergoes alterations. This transformation, coupled with favorable mutations and functional redundancy of core clock genes, bestows new physiological and metabolic advantages upon the resulting hybrids [29, 42].

Although histone modification is a critical factor affecting gene expression activity, gene expression regulation is influenced by multiple factors. Therefore, in our study, we performed chromatin immunoprecipitation experiments only on the histone modifications H3K4me3 and H3K9ac, which have been previously reported to be most strongly associated with diurnal oscillations [7, 11, 12, 43]. For alignment of the library fragments to the reference genome, we utilized the Burrow-Wheeler Aligner’s mem algorithm module [47]. Fragments with low alignment quality were filtered out from the library using the view module of SAMtools [48]. A script was utilized to call the view module of Sambamba software and retain uniquely aligned library fragments [45]. Duplicate sequences were removed from the library using Picard’s MarkDuplicates module. An index of the alignment result files was created using the index module of SAMtools [48]. For the calculation of small-volume alignment results files, we used the bamCoverage module of deepTools [49] and visualized the data with the Integrative Genomics Viewer [50]. The multiBamSummary module of deepTools was used to count differences between datasets [49]. To detect signals, we utilized MACS2 callpeak [51]. RNAPII recruitment has a broad peak shape, while H3K4me3 and H3K9ac have narrow peaks. To annotate peaks, we used the annotatePeaks.pl script from Homer [52]. To create annotation files for RNAPII, H3K4me3, and H3K9ac genome-wide peaks, we employed the intersect module of deepTools and a script [49]. The peak signal intensity was counted based on the peak annotation file using Htseq’s htseq-count module [53].

ATAC-seq and data analysis

We infiltrated the sample with 1× PBS solution (including 1× PI) in a small petri dish and homogenized it by repeatedly cutting it with a blade. The sample was then filtered twice through a single-layer membrane, and 20ml of the filtrate was collected. We used the nonfixed angle accessory of a sizeable refrigerated centrifuge to centrifuge the sample at 1000 G for 5 min at 4 °C. After removing the supernatant, we washed the pellet twice with 1× PBS solution (containing 1× PI) and then resuspended it in 10 ml of 1× PBS solution (containing 1× PI). Next, we washed the pellet once more with Transposase Incubation Buffer, and after each wash, we centrifuged it at 600G for 3 min at 4 °C and discarded the supernatant. (Transposase Incubation Buffer: 0.02M HEPES Buffer (pH 7.5), 0.3M NaCl, 0.5 µM Spermidine, 0.1 M MgCl2). To continue, we took an appropriate amount of cell nuclei and resuspended them in 500 µL of Tagmentation Buffer. The resuspended cell nuclei were incubated at a low temperature for 10 min and then treated in a shaking dry bath device at 37 °C for 30 min at 300 r/min. (Tagmentation Buffer: 492.5 µL Transposase Incubation Buffer, 7.5 µL 20% Triton X-100 containing 0.1 M MgCl2). DNA was purified using the QIAGEN MinElute PCR Purification Kit. The purified DNA was then prepared using the NEB Next Ultra II DNA Library Prep Kit for Illumina kit, and the PCR products were sorted by AMPure XP magnetic beads. The concentrations of the mixed magnetic beads were 0.55× and 0.2×, and the resulting library was subjected to PE150 sequencing using Illumina HiSeq X Ten.

We utilized the FastQC software tool to assess the data characteristics of the library. Subsequently, we employed Trimmomatic software tools to eliminate linkers and low-quality sequences from the library [43]. Burrow-Wheeler Aligner’s mem algorithm module aligned the library fragments to the reference genome [47]. To remove poorly aligned fragments, we used the view module of SAMtools [48], while uniquely aligned library fragments were retained using a script and by calling the view module of Sambamba software [45]. Duplicate sequences were removed from the library using Picard’s markduplicates module. We employed the index module of SAMtools to generate an index of the alignment result files [48]. Furthermore, we utilized the bamCoverage module of deepTools to compute alignment result files of small volumes, and visualized them using the Integrative Genomics Viewer [50]. MACS2 callpeak was employed to detect signals [51], whereas high-confidence peaks for technical replicates were obtained using IDR [54].

Correlation analysis of biological replicates

The similarity matrix between biological replicates was calculated using the multiBamSummary module of deepTools [49]. Subsequently, a heatmap or scatter plot was generated using the plotCorrelation module to visualize the correlation between the replicates.

Time-series analysis of diurnal oscillation

The local database was initially trained using Biocycle, and subsequently, a deep neural network was employed to calculate a data matrix, resulting in the derivation of time-series features for each gene [34, 35, 55]. Transcriptome data were normalized to the expression levels of protein-coding genes using the TPMCalculator, which calculates the expression values in transcripts per million (TPM) [46]. Meanwhile, the peak data associated with epigenetic modifications were normalized by the deepTools software package using bins per million mapped reads (BPM) to reflect the signal intensity of each peak [49]. Finally, the merged data generated a time-series matrix.

Data visualization

The unique alignment result files from the ATAC-seq, ChIP-seq, and RNA-seq datasets were used to generate bigWig format files by applying the BPM calculation standard through the bamCoverage module of deepTools [49]. These bigWig files were then utilized for data visualization with IGV software [50].