Integrated multi-omics profiling reveals dynamic regulation of light-induced chloroplast biogenesis in Brassica napus seedlings

Abstract

Photosynthetic efficiency, a pivotal determinant of crop yield, is governed by chloroplast development—a process that remains poorly understood in polyploid crops. Using tetraploid oilseed rape (Brassica napus) as a model, we phenotypically characterized chloroplast development under light induction and performed a high-resolution, multi-omics analysis of this process. Through the integration of time-series transcriptome, proteome, and post-translational modification (PTM) data—encompassing acetylation, phosphorylation, and ubiquitylation—we reveal a multi-layered regulatory network coordinating chloroplast maturation. A core, sequential transcription factor cascade orchestrates the temporal program, which is finely modulated by crosstalk between alternative splicing and PTMs. PTMs further fine-tune the activity of proteins within essential photosynthetic pathways. We also demonstrate differential subfunctionalization of homeologous gene pairs, a polyploid-specific strategy that enhances regulatory flexibility and robustness. Our findings establish a molecular map of chloroplast development, elucidating how transcriptional, post-transcriptional, and post-translational layers may contribute to efficient plastid maturation. This study also identifies upstream regulators, particularly within the photosystem and chlorophyll biosynthesis pathways, as potential candidates for functional validation to assess their roles in improving photosynthetic performance. Collectively, our findings provide a resource for future research in chloroplast biology, photosynthesis, polyploid biology, and comparative-omics studies.

Graphical Abstract

Graphical Abstract.

Introduction

Chloroplasts are semi-autonomous organelles where photosynthesis occurs, converting light energy into chemical energy that sustains plant metabolism and global ecosystems. Beyond energy transformation, they serve as metabolic hubs for producing essential compounds and signaling molecules, thereby integrating environmental cues with plant growth and stress responses [1]. The biogenesis of functional chloroplasts from proplastids requires coordinated activity between nuclear and chloroplast genetic systems. Given the limited coding capacity of the chloroplast genome, thousands of nuclear-encoded proteins must be imported, necessitating precise spatiotemporal organization for the proper assembly of photosynthetic complexes [2, 3]. This organziation involves multiple regulatory layers, including transcription, post-transcriptional processing, translation, and post-translational modifications (PTMs). However, a systematic, time-resolved understanding of how these layers contribute to drive chloroplast maturation remains incomplete. In particular, the molecular cascade linking light perception to proteome establishment, spanning from transcriptional reprogramming to protein accumulation, requires further elucidation through time-resolved multi-omics approaches.

Photosynthetic establishment is central to chloroplast function and involves the coordinated activation of pigment biosynthesis, light reactions, and carbon fixation pathways [4, 5]. Light signaling initiates the transition from skotomorphogenesis to photomorphogenesis, resulting in transcriptional reprogramming that promotes greening [6]. During light reactions, photosystems II and I (PSII, PSI) operate in concert to harvest light, split water, and generate ATP and NADPH while establishing proton gradients [7]. These processes are modulated by factors such as antenna composition and photoprotective mechanisms [8]. The Calvin–Benson cycle in the stroma then utilizes this energy for CO₂ fixation, primarily catalyzed by ribulose-1,5-bisphosphate carboxylase/oxygenase (Rubisco) [9]. The efficiency of carbon fixation is influenced by Rubisco’s catalytic constraints and the regeneration rate of its substrate, regulated by enzymes such as SBPase [10]. The entire process, from complex assembly to enzymatic activity, is regulated through multiple layers, including PTMs and stromal conditions [11].

Light is perceived by photoreceptors including phytochromes and cryptochromes, initiating a signaling cascade that regulates gene expression [12, 13]. In darkness, PHYTOCHROME INTERACTING FACTORS (PIFs) promote skotomorphogenesis and repress chlorophyll biosynthesis. Upon light exposure, activated photoreceptors promote PIF degradation and stabilize transcription factors (TFs) such as ELONGATED HYPOCOTYL5 (HY5), activating photomorphogenic programs that include chloroplast development [5, 13]. Beyond transcription, PTMs, including phosphorylation, acetylation, and ubiquitination, rapidly modulate protein activity, stability, and interactions. Despite existing knowledge, the dynamic and integrated regulatory sequence from light perception through transcriptional activation, protein accumulation, and PTM-mediated refinement across defined time points remains to be fully characterized.

This regulatory landscape is further complicated in polyploid species. Brassica napus is an allotetraploid derived from hybridization between B. rapa and B. oleracea, containing distinct A and C subgenomes [14]. In allopolyploids, homoeologous genes often exhibit functional divergence, including differential expression and regulation, which may contribute to phenotypic plasticity and adaptive variation [15]. How such subgenome-divergent pattern operates during the dynamic process of chloroplast biogenesis, and whether homoeologs play partitioned roles across different regulatory layers, remains unclear. Investigating these aspects could provide mechanistic insights into photosynthetic regulation in polyploid crops.

To address these questions, we conducted a time-resolved multi-omics study of etiolated B. napus seedlings during light-induced greening. Our analysis integrates transcriptomic, proteomic, and PTM (phosphorylation, acetylation, ubiquitination) profiles to map the dynamic regulatory network underlying chloroplast biogenesis, with a focus on photosystem, the Calvin–Benson cycle, and chlorophyll synthesis. This work reveals a temporally decoupled regulatory cascade in which rapid transcriptional changes precede delayed proteomic responses. We further examine the distinct and coordinated roles of major PTMs and stage-specific alternative splicing (AS) in shaping the functional proteome. Additionally, we investigate the widespread subfunctionalization of photosynthesis-related homoeologous gene pairs between the A and C subgenomes, offering insights into the complexity of polyploid species. This study contributes to our understanding of polyploid plant biology and offers a resource for future research in chloroplast biology, photosynthesis, and crop photosynthetic performance improvement.

Materials and methods

Plant materials and growth conditions

The conventional semi-winter B. napus cultivar ZhongShuang11 (ZS11) was used in this study. Seeds were germinated and grown in complete darkness at 22°C for 4 days. The resulting etiolated seedlings were then exposed to white light for the indicated durations, with a light intensity of 120 μmol m⁻² s⁻¹. Cotyledons were harvested at the specified time points, immediately frozen in liquid nitrogen, and stored at −80°C for subsequent analysis. Etiolated seedlings collected at time zero without light exposure served as the control.

Plasmid construction

To generate constructs for the transient expression assay, the CDS fragment of BnGLK1, BnMYB44, and BnBBX15 genes were inserted into pUC18-eYFP [16] to obtain pUC18-BnGLK1-eYFP, pUC18-BnMYB44-eYFP, and pUC18-BnBBX15-eYFP, respectively. The 2.0-kb promoter of BnLhcb6-A08, BnLhcb6-C08, BnMDH1, and BnPETC were amplified and inserted into the pGreenII0800-LUC [16] to generate the pBnLhcb6-A08-LUC, pBnLhcb6-C08-LUC, pBnMDH1-LUC, and pBnPETC-LUC, respectively.

Transient expression assay

To assess the transcriptional activation activity of candidate TFs, transient dual-luciferase assays were performed in B. napus mesophyll protoplasts. Protoplasts were isolated from leaves of 3–4-week-old B. napus plants and transfected using a PEG-mediated method as previously described [17]. Reporter constructs containing the firefly luciferase (LUC) gene driven by target gene promoters (BnLHCB6, BnMDH1, or BnPETC) were co-transfected with effector constructs expressing YFP-tagged BnGLK1, BnMYB44, or BnBBX15 under the control of the CaMV 35S promoter. A 35S:Renilla luciferase (REN) construct was included in each transfection as an internal control (see Supplementary Table S1 for plasmid construction sequences). After overnight incubation, LUC and REN activities were measured using the Dual-Luciferase Reporter Assay System (Promega) according to the manufacturer’s instructions. Relative promoter activity was expressed as the LUC/REN ratio.

qRT-PCR analysis

Total RNA was extracted from seedlings at the indicated time points using an RNA Extraction Kit (Tiangen, Beijing, China) according to the manufacturer’s instructions. First-strand complementary DNA (cDNA) was synthesized from total RNA using M-MLV reverse transcriptase (Invitrogen). Quantitative reverse transcription polymerase chain reaction (qRT-PCR) was performed on a LightCycler 480 system (Roche) with SYBR Premix ExTaq (TaKaRa), following the respective manufacturers’ protocols. Gene-specific primers used in this study are listed in Supplementary Table S1. Each reaction was performed with technical replicates, and relative gene expression levels were normalized to the reference gene.

Chlorophyll quantification

Frozen tissues of B. napus cotyledons were extracted by the addition of 80% (v/v) acetone and measured according to previous reports [16].

Chlorophyll fluorescence measurements

Chlorophyll fluorescence parameters were measured using a HEXAGON-Imaging-PAM system (Zealquest Scientific Technology) following the manufacturer’s instructions. Brassica napus cotyledons were harvested at the indicated time points (0, 0.5, 2, 6, and 12 h) after transfer from darkness to continuous white light (120 μmol m⁻² s⁻¹). For Fv/Fm determination, cotyledons were dark-adapted using leaf clips for 15 min and then exposed to a saturating light pulse (5000 μmol m⁻² s⁻¹, 0.8 s). Minimal fluorescence (F₀) was recorded under weak measuring light (0.04 μmol m⁻² s⁻¹), and maximal fluorescence (F_m) was obtained during the saturating pulse. The maximum quantum yield of photosystem II was calculated as Fv/Fm = (F_m − F₀)/F_m. Measurements were performed on 15–20 cotyledons per replicate, with three independent biological replicates.

RNA isolation and RNA-seq library preparation

Total RNA was extracted using the CTAB-PBIOZOL method. RNA quality and concentration were assessed using a Qubit fluorometer and a Qsep400 high-throughput bioanalyzer. Strand-specific cDNA libraries were constructed using poly(A) messenger RNA (mRNA) enrichment, fragmentation, and reverse transcription with random hexamers. Second-strand synthesis incorporated dUTP to preserve strand orientation. Following end repair, A-tailing, and adapter ligation, libraries were size-selected (250–350 bp inserts), amplified by PCR, and validated for fragment size and concentration. Pooled libraries were sequenced on an Illumina NovaSeq X Plus platform to generate 150-bp paired-end reads via sequencing-by-synthesis.

RNA-seq data analysis

Raw reads were processed with fastp (v0.32.2) for adapter trimming and quality control using parameters:-n_base_limit 15, and qualified_quality_phred = 20 [18]. The resulting high-quality reads were aligned to the B. napus reference genome (ZS11.v0) using HISAT2 (v2.2.1) [19, 20]. Gene-level expression quantification was performed with featureCounts, and normalized values were calculated as fragments per kilobase of transcript per million mapped reads (FPKM) [21]. For transcript-level analysis, novel transcripts were assembled and predicted with StringTie (v2.1.6), and AS events were identified and compared across sample groups using rMATS (v4.1.2) [22, 23]. To evaluate data reliability, we conducted hierarchical clustering, and Pearson correlation analysis between biological replicates using R (v4.4.1). Differentially expressed genes (DEGs) were identified with DESeq2 (v1.22.1) [24, 25]. Genes exhibiting an absolute log₂-fold change (|log₂FC|) ≥ 1 and a false discovery rate (FDR) < 0.05 were considered statistically significant. Gene Ontology (GO) enrichment analysis was performed using the enrichGO function from the clusterProfiler package [26, 27]. Concurrently, Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis was performed using the KEGG database (https://www.genome.jp/kegg) [28]. Statistical significance was assessed with a hypergeometric test, and results were visualized using dot plots generated in R. Summary statistics for the RNA-seq dataset are provided in Supplementary Dataset 1.

Protein extraction and preparation for mass spectrometry quantification

Cotyledon samples were fixed in 2.5% glutaraldehyde in phosphate buffer at 4°C for 4 h. Following fixation, the tissues were rinsed and subsequently post-fixed in 1% osmium tetroxide (OsO₄) at 4°C overnight. After thorough rinsing with phosphate buffer, the samples were dehydrated through a graded ethanol series. They were then infiltrated with a graded series of epoxy resin in propylene oxide and finally embedded in Spurr’s resin (Sigma–Aldrich). Ultrathin sections were cut using a Reichert OM2 ultramicrotome, collected on single-slot copper grids, and stained with uranyl acetate followed by lead citrate. Sections were observed and imaged under a JEM-1230 transmission electron microscope (JEOL).

Liquid chromatography was performed on a nanoElute UHPLC system (Bruker Daltonics) using a reverse-phase C18 column (25 cm × 75 μm, 1.6 μm, IonOpticks) at 50°C. Peptides were separated over a 60-min gradient from 2% to 35% mobile phase B (0.1% formic acid in acetonitrile) at a flow rate of 0.3 μl/min.

Mass spectrometry (MS) data were acquired on a timsTOF Pro2 (Bruker Daltonics) coupled online via a CaptiveSpray source, operated in PASEF mode. Full MS scans (100–1700 m/z) were followed by MS/MS scans of the top 10 precursors per cycle. Ion mobility was set between 0.7 and 1.4 Vs/cm². Collision energy varied linearly with mobility.

Proteomic data analysis

MS raw data from data-independent acquisition (DIA) experiments were analyzed using DIA-NN (version 1.8.1) in library-free mode [28]. A spectral library was constructed de novo from the DIA data using the software’s deep learning-based neural network algorithms. This library generation incorporated two protein sequence databases: (i) the reference genome protein sequences for ZS11.v0 (100 919 sequences), and (ii) novel protein sequences predicted from our RNA-Seq data (4541 sequences). The Match Between Runs (MBR) feature was enabled during this process to maximize identifications. Following initial library creation, the DIA data were re-searched against this project-specific spectral library for final quantification. Protein quantification was performed using the MaxLFQ algorithm integrated within DIA-NN. Subsequently, protein intensity values were extracted from the results. Relative protein (R) was determined by performing within-sample normalization on the raw intensity data. For each sample, the intensity of each feature (protein, j) was devided by the median intensity of all features within that same sample (i), using the formula: R_ij = I_ij / Median (I_i). To further evaluate data reliability, we conducted principal component analysis (PCA) and Pearson correlation analysis between biological replicates using R (v4.4.1). Proteins were defined as differentially expressed (DEPs) based on between-sample comparison with thresholds of a fold change ≥ 1.5 (increased accumulation) or ≤ 0.67 (decreased accumulation) and a P-value ≤ .05. Finally, the genes corresponding to these differentially expressed proteins were functionally characterized, beginning with prediction of subcellular localization using WoLF PSORT software [29]. GO enrichment analysis was performed using the enrichGO function from the clusterProfiler package [26, 27]. Concurrently, KEGG pathway enrichment analysis was performed using the KEGG database (https://www.genome.jp/kegg) [28]. Statistical significance was assessed with a hypergeometric test, and results were visualized using dot plots generated in R. Summary statistics for the proteomic dataset are provided in Supplementary Dataset 2.

Modified peptide preparation, enrichment, and library preparation

Frozen plant tissues were ground to a fine powder in liquid nitrogen. An appropriate amount of powder was transferred to a 1.5 ml tube and homogenized in L3 lysis buffer (1% sodium dodecyl sulfate, 100 mM Tris–HCl, 7 M urea, 2 M thiourea, 1 mM phenylmethylsulfonyl fluoride, 2 mM ethylenediaminetetraacetic acid) supplemented with 1% PhosSTOP™ (Sigma). The mixture was sonicated on ice for 10 min and centrifuged to collect the supernatant. Proteins were precipitated overnight at –20°C by adding four volumes of cold acetone. The pellet was washed with cold acetone, dissolved in 8 M urea, and quantified using a BCA assay kit.

For each sample, 0.5 mg of protein was reduced with 10 mM dithiothreitol (37°C, 45 min) and alkylated with 50 mM iodoacetamide (room temperature, 15 min in the dark). Proteins were precipitated with cold acetone, washed, and digested overnight at 37°C with trypsin (Promega) and Lys-C (Thermo Fisher) in 25 mM ammonium bicarbonate. The digest was acidified with TFA to pH 2–3, centrifuged, and peptide concentration was determined using a Pierce™ Quantitative Peptide Assay Kit (Thermo Fisher).

For acetyl-, phospho-, and ubiquitin-peptide enrichment, peptides were dissolved in IMAC loading buffer (80% acetonitrile, 6% trifluoroacetic acid saturated with glutamic acid) and incubated with pre-washed IMAC beads with gentle agitation. Beads were washed sequentially with 50% acetonitrile/ 6% TFA/ 200 mM NaCl and 30% acetonitrile/ 0.1% TFA. Bound acetyl-, phospho-, and ubiquitin- peptides were eluted with 10% ammonia, vacuum-dried, desalted using C18 ZipTips, and quantified prior to LC-MS/MS analysis.

PTM-omics data analysis

MS raw data were analyzed with FragPipe (v21.1), which relies on MSFragger for qualitative analysis and uses Phosopher for validation and filtering [29]. Spectra files were searched against the combined.fasta database (a total of 105 461 sequences, including reference and novel protein sequences) concatenated with decoy and contaminants database using the following parameters: the mass tolerances of precursor ions and fragment ions were both set as 20 ppm; trypsin and Lys-C were specified as digestive enzyme allowing up to two missing cleavages; the peptide length range was set to 6–50 amino acid residues; carbamidomethyl on Cys was specified as fixed modification, and acetylation on protein N-terminal, phospho on Ser, Thr, and Tyr, and ubiquitination on Lys were specified as variable modifications with a maximum of three variable modifications; FDR of search results was adjusted to <1% at both protein and peptide levels, and deleted decoys and contaminants. For label-free quantitation, select IonQuant for the quantitative module, check MBR and MaxLFQ. For PTMs, the minimum site localization probability was set to >0.75. Proteins which contained more than one razor peptides were used for further quantification analysis. Relative PTM abundances (R) were determined by performing within-sample normalization on the raw intensity data. For each sample, the intensity of each feature (PTM site, j) was divided by the median intensity of all features within that same sample (i), using the formula: R_ij = I_ij / Median (I_i). To further evaluate data reliability, we conducted PCA, and Pearson correlation analysis between biological replicates using R (v4.4.1). PTM sites were defined as differentially expressed sites based on between-sample comparison with thresholds of a fold change ≥ 1.5 (increased accumulation) or ≤ 0.67 (decreased accumulation) and a P-value ≤ .05. Finally, the genes corresponding to these differentially expressed sites were functionally characterized, beginning with prediction of subcellar localization using WoLF PSORT software [30]. GO enrichment analysis was performed using the enrichGO function from the clusterProfiler package [26, 27]. Concurrently, KEGG pathway enrichment analysis was performed using the KEGG database (https://www.genome.jp/kegg) [28]. Statistical significance was assessed with a hypergeometric test, and results were visualized using dot plots. Clustered heatmaps were plotted with ComplexHeatmap (v2.12.0). Summary statistics for the PTM dataset are provided in Supplementary Dataset 3 for acetylation, Supplementary Dataset 4 for phosphorylation, and Supplementary Dataset 5 for ubiquitination.

Gene regulatory network analysis

All DEGs had their expression values scaled to a 0–1 range. Subsequently, K-means clustering analysis was performed using the following parameters: iter.max = 100, centers = 10, and nstart = 25. This resulted in the grouping of genes into ten distinct clusters. The optimal number of clusters was determined using the Gap Statistic method, implemented via cluster (v2.1.8). For each cluster, TFs were identified using iTAK (v1.7a) [31]. Their frequency distribution was visualized using a word cloud generated by wordcloud2 (v2_0.2.1). Subsequently, gene regulatory networks (GRNs) were inferred for each cluster using GENIE3 (v1.28.0) [32] with the following parameters: regulators set as both transcription regulators (TRs) and TFs, and nTrees = 500. A link weight cutoff of 0.05 was applied. The resulting networks were constructed and visualized using the visNetwork (v2.1.4). Within each visualized network, TFs and non-TFs were distinguished, and the 10 TFs with the highest number of connections were highlighted and labeled.

Protein–protein interaction network analysis

For all differentially accumulated molecules, protein–protein interaction (PPI) networks were constructed by diamond using the STRING database (http://string-db.org/) [33]. For the PPI analysis, interaction relationships were retrieved with a confidence score threshold of >400.

Upstream regulator identification

For all pathway enzyme genes involved in chlorophyll synthesis, photosynthesis, and the Calvin cycle, we extracted their connections to transcription regulators from both the co-expression GRNs and PPIs data. After removing duplicate connections identified across different time points, we calculated the number of connections for each regulator and ranked them accordingly.

Homoeologous gene pairs identification

Homoeologous gene pairs between the A and C subgenomes were identified using SynOrths [34]. Their putative orthologs in Arabidopsis thaliana were determined based on annotations from the BnIR database [20].

Regarding homoeolog naming, homoeologs are designated by the prefix “A-” or “C-” followed by the chromosome information and then the gene name. This indicates both their subgenome of origin in this tetraploid species and the chromosome location (e.g. A08.COP1, where “A08” denotes chromosome A08). Additionally, a comprehensive glossary table of all abbreviations used in the text is provided as Supplementary Table S2.

Results

Kinetic and ultrastructural analysis of the etioplast-to-chloroplast transition in B. napus cotyledons

To investigate the etioplast-to-chloroplast transition in dark-grown cotyledons of B. napus, we exposed etiolated seedlings to continuous light and collected samples at five time points (0, 0.5, 2, 6, and 12 h after illumination) for phenotypic, cytological, and biochemical characterization (Fig. 1A). We then quantified the greening progression and measured chlorophyll content. In dark-grown cotyledons (0 h), chlorophyll levels were low, consistent with an etiolated state. Upon illumination, chlorophyll accumulation began to increase slightly between 0.5 and 2 h, followed by a sharp rise by 6 and 12 h, which corresponds well with the observed greening phenotype (Fig. 1A and Supplementary Fig. S2A).

Figure 1.

Experimental design and key datasets. (A) Temporal dynamics of light-induced greening in rapeseed seedlings under different light exposure durations. (B) Phenotypic changes of rapeseed chloroplasts across varying light durations, scale bars = 500 nm. (C) Photosystem II maximum quantum efficiency (Fv/Fm) during light-induced greening in B. napus cotyledons. Fv/Fm values were measured on dark-adapted cotyledons at the indicated time points after transfer to continuous white light (120 μmol m⁻² s⁻¹). Data are presented as mean ± SD from three independent biological replicates. (D) Schematic workflow of multi-omics analysis (transcriptome, proteome, and PTM profiling) during the early stage of chloroplast development.

Transmission electron microscopy was used to monitor plastid ultrastructure during greening. In dark-adapted cotyledons, plastids exhibited the characteristic morphology of etioplasts, containing one or more semi-crystalline prolamellar bodies (PLBs) comprised of a tubular network. After only 0.5 h of illumination, a clear transformation of the PLB was evident. The highly ordered paracrystalline lattice became irregular and less compact, indicating the initiation of PLB disassembly. By 2 h of light, PLB structures were further reduced in plastid sections. Concurrently, an increase in the number of interconnected thylakoid membranes (lamellae) was observed, indicating the transition from PLB-derived prothylakoids to the early thylakoid network. At 6 h, the thylakoid membrane system underwent substantial expansion and organization, with a notable increase in the number and size of grana stacks (appressed thylakoids) and the appearance of initial starch granules in the stroma. After 12 h of illumination, plastids exhibited further maturation, resembling young chloroplasts. Grana stacks were more numerous and exhibited greater stacking (Fig. 1B). To assess whether this structural progression is accompanied by the acquisition of photosynthetic competence, we measured the maximum quantum efficiency of photosystem II (Fv/Fm) over the same time course. Fv/Fm values remained extremely low during the first 0.5 h of illumination, confirming that PSII was functionally immature and that light energy conversion efficiency was negligible. Detectable PSII activity emerged by ~2 h and increased steadily at 6 and 12 h (Fig. 1C). The gradual rise in Fv/Fm coincides temporally with the expansion of thylakoid membranes and the formation of grana stacks described earlier. Together, these observations reveal a rapid and orderly progression from etioplasts to developing chloroplasts, with kinetics largely consistent with those reported for tobacco and other dicots [4].

To investigate the molecular mechanism underlying light-induced chloroplast biogenesis, we employed a multi-omics strategy. This approach leveraged five newly generated, high-quality datasets: transcriptomic, proteomic, and three post-translational omics types across five time points (Fig. 1D). This design captures the system’s dynamics and facilitates from gene expression to functional protein activity.

Integrated multi-omics reveals a temporally decoupled transcriptional and translational cascade during chloroplast biogenesis

Time-resolved transcriptomic and proteomic profiling of seedlings exposed to light (0, 0.5, 2, 6, and 12 h) were performed to obtain a systems-level understanding of chloroplast biogenesis. Transcriptome analysis identified a total of 44 185 expressed genes (FPKM ≥ 1). Among these, 2337 were novel transcripts not previously annotated in the reference genome. Analysis of the proteome identified a total of 12 817 proteins, of which 301 were novel and not previously annotated in the reference genome. The Pearson correlation coefficients between replicate samples were high for both the RNA-seq and proteomic data (r ≥ 0.97, Supplementary Fig. S1A), indicating strong consistency. Hierarchical clustering of the RNA-seq data showed replicates clustering together (Supplementary Fig. S1B), and PCA of the proteomic data grouped samples clearly by treatment (Supplementary Fig. S1C). These analyses confirm the high reproducibility of the datasets. qRT-PCR validation further confirmed that the expression patterns of three BnLHCB genes matched their transcriptomic profiles (Supplementary Fig. S2B and C).

To further investigate the molecular mechanisms underlying the differential responses to varying durations of light exposure, we conducted pairwise differential expression analysis between samples from each time point (0.5, 2, 6, and 12 h) and the initial dark-grown state 0 h. In total, 15 028 unique genes were defined as DEGs, illustrating the scale of this environmental response. Upon illumination, the transcriptome underwent immediate and extensive rewiring, with the number of DEGs increasing from 5281 at 0.5 h to a peak of 8824 at 6 h before a slight decline at 12 h (Fig. 2A and Supplementary Fig. S2G). In stark contrast, the proteome responded with a significant temporal lag; 1511 unique proteins were identified as differentially expressed proteins. Protein-level changes accumulated progressively, with only 217 DEPs at 0.5 h, not reaching their peak (1048) until 12 h (Fig. 2A and Supplementary Fig. S2H). This delay is exemplified by the BnLHCB family, where protein accumulation lagged well behind transcript induction (Supplementary Fig. S2C–E). This kinetic disconnect establishes that light-induced chloroplast development is paced by a primary transcriptional wave that precedes and instructs a slower translational output.

Figure 2.

Comprehensive multi-omics analysis. (A) Sankey diagram illustrating the dynamic accumulation of gene transcripts (Tr) and their corresponding proteins (Pr). Upper (green) streams represent differentially accumulated molecules, while lower (gray) streams represent non-differential accumulation. Differential expression thresholds: Transcripts (DE Tr), |log₂FC| ≥ 1 and q-value ≤ 0.05; Proteins (DE Pr), FC ≥ 1.5 or ≤ 0.67 and P-value ≤ .05. (B) GO enrichment analysis of DEGs. (C) GO enrichment analysis of DEPs. (D) Venn diagram of all differentially expressed molecules across the multi-omics datasets, including DEGs, DEPs, differentially acetylated proteins (DEAs), differentially phosphorylated proteins (DEPPs), and differentially ubiquitinated proteins (DEUs). (E) Subcellular localization distribution of the differentially accumulated molecules, including proteins, acetylated proteins, phosphorylated proteins and ubiquitinated proteins.

Then, we performed GO enrichment analysis to characterize the function of these genes. The results revealed a coherent, stage-specific transcriptional progression. Early responses (0.5 h) were restricted to light perception and signaling. By 2 h, the program expanded explosively to encompass photosynthesis and light-harvesting and reaction (Fig. 2B). Later stages (6, 12 h) were dominated by transcripts encoding structural components of the thylakoid and lumen, indicating a shift to morphogenesis. Proteomic enrichment told a congruent but delayed pattern. The earliest protein signature (0.5 h) was not light-specific but related to translation (ribosomal proteins), suggesting a wholesale ramp-up of protein synthesis capacity. The corresponding photosynthetic transcripts detected at 2 h were manifest as significantly accumulating proteins only from 2 h onward, with levels rising through 12 h (Fig. 2C). Furthermore, KEGG pathway enrichment analysis of DEGs outlined a metabolic trajectory from core energy production to anabolism: photosynthesis pathways were enriched by 2 h, carbon fixation by 6 h (Supplementary Fig. S4A). Proteomic data confirmed this sequence but also suggested potential post-transcriptional regulatory nodes. While Calvin cycle enzyme accumulation aligned with their transcript timing (6 h), the accumulation of photosynthesis and antenna proteins was disproportionately delayed. Furthermore, the enzymatic machinery for starch and sucrose metabolism appeared only at the proteome level at 12 h, despite earlier transcriptional activation (Supplementary Fig. S4B). This indicates that the transition to photosynthetic autotrophy and carbon storage may involve both transcriptional and post-transcriptional levels accumulation.

Transcriptional regulatory network governing chloroplast development in rapeseed

PCA of transcriptome showed a clear developmental trajectory directly correlating with the time of light exposure (Fig. 3A). The initial separation, driven directly primarily by PC2, distinguished the 0 and 0.5 h samples. Subsequently, the 2 h time point diverged markedly along both PC1 and PC2—a pattern consistent with hierarchical clustering analysis (Supplementary Fig. S1B). By the later stages, the transcriptional program appeared to stabilize, evidenced by the close clustering of the 6 and 12 h samples, which suggests the completion of the major light-induced developmental transition, a process that corresponds with cotyledon greening.

Figure 3.

Transcriptional regulatory network governing chloroplast development. (A) The developmental time unites are defined by PCA. (B) K-means clustering of DEGs from four time points (0.5, 2, 6, and 12 h), showing the number of genes for each cluster. A yellow–green color scale was used, where green indicates high accumulation. (C) Enriched TF families and GO terms for each cluster in panel (B). A continuous scale was used, where a longer bar indicates a higher probability of enrichment. (D) GRNs for each cluster, with the top 10 TF hub genes highlighted in the center. In the networks, TFs are shown in green and non-TFs in gray.

To identify the regulators that drive genome-wide transcriptional dynamics in response to varying durations of light exposure, we first performed k-means clustering on all DEGs, which resolved the transcriptomic response into ten distinct co-expression clusters (Fig. 3B and Supplementary Fig. S3A). Notably, five clusters exhibited pronounced time-specific expression patterns, revealing a phased transcriptional program. Mirroring this progression, TF activity shifted dynamically: AP2/ERF-ERF family were predominant during the early response (0–0.5 h), followed by a surge in bHLH at 2 h, and a subsequent dominance of MYB from 6 to 12 h (Fig. 3B and C). Further GO enrichment analysis of these genes in time-specific clusters provided functional context for these transcriptional phases. The early, uninduced state (C10, 0 h) was associated with baseline developmental processes such as organ morphogenesis. Subsequent clusters were linked to protein folding (C2, 0.5 h) and direct response to light stimuli (C1, 2 h). Finally, the later-phase clusters, C4 (6 h) and C9 (12 h), were enriched for functions related to the establishment of photosynthetic capacity, including photosynthesis and carbon fixation, respectively (Fig. 3C). Taken together, this finding suggests a relatively tight, time-dependent trajectory of gene expression.

While five time-specific clusters defined a clear phased program, the remaining clusters exhibited co-expression patterns spanning multiple time points. These extended clusters revealed distinct categories of transcriptional regulation: sustained repression (C3), delayed activation (C5), progressive repression (C6), transient inhibition (C7), and a dedicated transitional state (C8) (Supplementary Fig. S3A). Notably, the dynamics of these extended clusters were aligned with a sequential progression of TF family activity—from early AP2/ERF-ERF (C6, C8) to peaking bHLH (C5) and later C2H2 (C7) and C2C2-CO-like (C3) (Fig. 3C and Supplementary Fig. S3B). Together with contributions from lower-frequency families, this sequential cascade of high-frequency TFs appears to orchestrate a precise developmental sequence by fine-tuning gene expression.

To further explore the connectivity, we reconstructed high-confidence, stage-specific GRNs from time-series data. For each of the ten developmental clusters (C1–C10), we identified the top ten TFs with the highest network centrality, defining core regulatory hubs for each stage (Fig. 3D and Supplementary Fig. S3C). A comparative analysis of network topology revealed a dramatic architectural shift across the developmental timeline (Fig. 3D). The pre-stimulation network (C10), despite encompassing the largest gene set, was remarkably sparse—a simple architecture with few TFs exhibiting limited connectivity, indicative of a transcriptionally repressed or poised state. Light exposure triggered rapid and dramatic restructuring, marked by the sequential recruitment of new TFs into dominant hub positions (e.g. in C1–C8). This process drove an explosive increase in TF-target connections to downstream structural genes, culminating in the establishment of a dense GRN by the final stage (C9). Of the 10 distinct co-expression modules identified, C1 and C8 are the primary candidates for driving the transition to photomorphogenesis due to their tight association with cotyledon greening. The C8 module is defined by light-signaling hubs (e.g. ABI4, SPL15, BBX19), while C1 is populated by developmental executors (e.g. HYH, TCP10, TT8). These genes are well-known participants in chlorophyll biogenesis (HYH) or have been implicated in related light induced and pigment pathways with potential functions in this process (e.g. ABI4, BBX19, SPL15, TT8) [35–38].

Consequently, TFs-centered GRN analysis moves beyond static interaction maps to delineate a temporally ordered transcriptional cascade. This cascade appears to translate an external light signal into the gene expression associated with chloroplast biogenesis.

A multi-layered post-translational landscape directs chloroplast development

To further capture the functional layer beyond transcription, we performed a comprehensive profiling of three PTMs: acetylation, phosphorylation, and ubiquitination. A total of 8006 proteins (including 513 novel proteins) corresponded to 38 291 modification sites overall. Modifications were distributed as follows: PTM-A in 3071 proteins, PTM-P in 6717 proteins, and PTM-U in 4568 proteins, highlighting a vast and underexplored modulatory layer in chloroplast biogenesis. Data reproducibility was high, with an average inter-replicate Pearson correlation coefficient exceeding 0.9 (Supplementary Fig. S1A). Furthermore, PCA results showed group separation for the PTM sample data (Supplementary Fig. S1D–F).

At the transcript and protein levels, the cellular response is characterized by a greater number of genes showing increased accumulation than decreased accumulation (Supplementary Fig. S1G and H), suggesting a broad, activation-oriented program. However, this picture is refined by a more dynamic and opposing layer of modulation at the level of PTMs (Supplementary Fig. S1I–K). Specifically, acetylation shifts from an initial bias toward inhibition to a later balanced state. In contrast, phosphorylation is consistently associated with increased accumulation, while ubiquitination shows the opposite trend, favoring increased protein turnover. This indicates that the net cellular outcome is not dictated simply by transcript and protein abundance, but is precisely modulated by counterbalancing PTM signals—with phosphorylation providing “on” switches and ubiquitination acting as “off” switches. Beyond this, the protein targets of each PTM are dynamic and stage-specific (Supplementary Fig. S4C), indicating sophisticated, time-partitioned organization. Light therefore triggers not just a transcriptional cascade, but a complex post-translational network. Cellular localization analysis revealed that this differentially accumulated proteome, as well as the differentially accumulated PTM variants (PTM-A, PTM-P, and PTM-U), was predominantly distributed (73.27%–86.58%) across the nucleus, chloroplast, and cytoplasm. Further analysis showed an increasing localization preference for the chloroplast over time, suggesting a functional link between these proteins and chloroplast-related processes. Specifically, proteins with differential acetylation (DEAs) were primarily localized to the chloroplast and cytoplasm, while those with differential phosphorylation (DEPPs) were mainly nuclear, and differentially ubiquitinated (DEUs) proteins were distributed evenly across all three compartments. Notably, although the number increased over time (except for PTM-U), their proportional distribution among these compartments remained consistent (Fig. 2E). Collectively, these patterns suggest compartment-specific activities functions for each PTM during chloroplast development.

Next, to investigate the concordance of molecules across transcriptomic, proteomic, and PTM datasets during chloroplast development, we combined the differentially accumulated molecules from all time points, enabling an analysis that is not confounded by temporal delays in expression. The results revealed that a high proportion of molecules within each omics dataset were uniquely differentially accumulated and associated with chloroplast development. Specifically, these included 88.31% of DEGs, 39.44% of DEPs, 51.54% of DEAs, 63.24% of DEPPs, and 47.54% of DEUs (Fig. 2D). This finding suggests that abundant post-transcriptional and post-translational regulation are associated with chloroplast development. Further analysis of the subcellular localization of these unique proteins revealed that 20%–30% of the proteins and PTMs were located in the chloroplast (Supplementary Fig. S4D). Functional enrichment analysis of the uniquely associated proteins supported this organelle-specific signature, with pathways directly involved in chloroplast development being significantly overrepresented (Supplementary Dataset 2). Interestingly, phosphorylated proteins were distinctly enriched in RNA AS-related pathways (Supplementary Dataset 4). This finding aligns with their predominant nuclear localization and suggests a regulatory axis where nuclear-localized phosphorylation events may influence chloroplast development through the modulation of RNA processing.

Taken together, these results suggest that gene expression, protein accumulation, and PTMs function in a complementary and multi-layered manner to orchestrate chloroplast development.

Alternative splicing dynamics exhibit temporal specificity and link to chloroplast pathways

We profiled the landscape of AS during chloroplast development. AS has been reported to fine-tune gene expression in response to environmental stimuli during developmental processes [39]. During light-induced chloroplast development, the main AS events were alternative 3′ splice site (A3′SS) and retained intron (RI) (Fig. 4A). This matches earlier findings in plant environmental signaling studies [40, 41]. The landscape of AS was highly dynamic. The number of AS genes shared in pairwise comparisons changed considerably over the course of light exposure. Notably, 57% (2586/4542) of AS-associated genes were unique to specific time points, while <0.1% showed AS across all time points (Fig. 4B). This temporal specificity suggests that AS may serve as a rapid, flexible mechanism to adjust the proteome at defined developmental stages. At the single-gene level, we observed dynamic shifts in the expression of two major isoforms during the light exposure period (Fig. 4C), further underscoring the precision of this regulation.

Figure 4.

The AS events are associated with chloroplast development. (A) Number of genes undergoing AS events at each time point of light exposure. (B) Upset plot (modified version) showing the number of shared genes involved in AS events across different pairwise comparisons. Numbers in the brackets show the total number of genes per comparison. (C) Example of a gene exhibiting two major isoform types (A5SS and RI) that dynamically change across light exposure time points. (D) The heatmap depicts genes undergoing AS events across different durations of light exposure. Color intensity represents the level of AS relative to the 0 h, where red indicates a high level, blue indicates a low level, and white indicates no significant change.

To further explore the role of AS in chloroplast development, we analyzed the splicing profiles of genes involved in transcripts encoding photoreceptors, chlorophyll biosynthesis enzymes, and core components of the Calvin cycle and photosystems (Fig. 4D). These AS events were categorized into four main types: A3′SS, alternative 5′ splice site (A5’SS), RI, and skipped exon (SE). Hierarchical clustering grouped the genes into two major clusters according to their relative changes in AS compared to the dark-grown samples: one with reduced AS frequency and one with increased AS frequency. Each cluster could also be subdivided according to their temporal expression patterns. For example, genes such as HEMF1 (Heme ferrochelatase 1, functioning in tetrapyrrole biosynthesis), the histone deacetylase HDA9 (Histone deacetylase 9, a nuclear regulator mediating retrograde signal-responsive gene expression), the light signaling component SPA3 (Suppressor of PhyA-105 3, which links nuclear photoreceptor signaling to chloroplast maturation), and protochlorophyllide oxidoreductase (POR) all exhibited decreased AS in light relative to the dark condition. In contrast, genes including PORB (a chloroplast-localized enzyme essential for chlorophyll synthesis), the chloroplast inner envelope translocon subunit TIC110 (translocon at the Inner envelope membrane of chloroplasts 110 kDa) transduction, the photosystem I subunit PSAN (photosystem I subunit N), the uroporphyrinogen III decarboxylase, HEMG1 (Heme G1), and the chaperonin cpn60 all showed increased AS level. Crucially, this observed changes splicing dynamics directly shaped the isoform repertoire of genes essential for chloroplast biogenesis. It suggests that AS acts may not merely as a passive process, but also an active modulator of the core functional machinery required for chloroplast biogenesis and photosynthetic capacity. Together, these results provide new evidence on the role of AS in coordinating nucleus-chloroplast signaling crosstalk during chloroplast development.

Subfunctionalization of homeologous genes orchestrates chloroplast development in tetraploid B. napus

The allopolyploid genome of B. napus, derived from hybridization between B. rapa (AA) and B. oleracea (CC), provides a rich source of genetic material through the expansion and retention of duplicated gene families. This is particularly evident for genes enabling photosynthesis and chloroplast development, a duplication that may underpin the enhanced photosynthetic capacity of B. napus compared to diploid relatives like A. thaliana. To quantify this, we performed a homology analysis of 221 Arabidopsis genes involved in these processes. This revealed pervasive duplication, with most orthologs present as multiple copies—predominantly two or four, yet with several genes expanded beyond 10 copies (Supplementary Fig. S5A). This extensive redundancy presents a potential for complex regulation. We therefore focused on genes encoding BBX TFs, light receptors, and their downstream regulators, observing three expression patterns: (i) temporal partitioning, where initially co-expressed homeologs like B-Box domain protein BBX4 (A04/C04) diverged over time; (ii) clear subgenome bias, with dominant expression of C-subgenome copies in families like BBX25 (C04.BBX25) but A-subgenome copies in photoreceptors like PHYA (A06.PHYA); and (iii) complex family-specific coordination, where dominance was pair-specific (e.g. HYH family), involved consistent co-expression of specific pairs (e.g. A10/C09.PHOT2), or showed subgenome-wide bias across multiple pairs (e.g. C-subgenome dominance in PIF4, phytochrome interacting factor 4) (Supplementary Fig. S5B–L).

At the protein level, subfunctionalization was also observed (Supplementary Fig. S6A). This pattern is evident in the group-specific dominance of protein accumulation, as seen with C09.PHOT2, A08.CRY2, C05.PHY, and C09.HY5, and extends to PTMs such as acetylation, phosphorylation, and ubiquitination (Supplementary Fig. S6B–D). Specific examples of this group-specific pattern include acetylation (e.g. A07.BBX15, A06.PHOT1, A06.PHYA, A01.PHYE), phosphorylation (e.g. C03.PHOT1, C09.PHOT2, A08.CRY2, A01.HY5, A02.HY5), and ubiquitination (e.g. C03.PHOT1, A08.CRY2, A08.PHYA, A05.PHYB). The acetylation patterns of the blue light receptor phototropin 1 (PHOT1) and phytochrome A (phyA) align with those reported in Arabidopsis [42, 43]. The timing of ubiquitination accumulation on photoreceptor genes—after light illumination—is consistent with previous reports, occurring at multiple sites with a dynamic profile (Supplementary Fig. S6D). It adds an additional layer of complexity to the modification landscape during chloroplast development.

To further investigate the expression dynamics of the A and C subgenomes, we classified genes based on their expression levels as A- or C-dominated (Supplementary Fig. S7A). We observed a genome-wide balance, with only a minor proportion (∼5%) of genes exhibiting clear A- or C-subgenome dominance (Supplementary Fig. S7B). This finding supports earlier studies in B. napus and differs sharply from the lasting dominance seen in polyploids like wheat [44, 45]. The proportion of subgenome-dominated genes doubled to ~10% among DEGs, within which A- and C-dominated genes were evenly distributed (Supplementary Fig. S7C). Statistical analysis confirmed that subgenome-dominated genes are significantly enriched in DEGs compared to the background of all expressed genes (Supplementary Fig. S7H). This indicates that while the genome maintains broad homoeolog balance, light or developmental signals that induce differential expression can transiently bias homoeolog usage.

Beyond transcriptional dynamics, we identified distinct and opposing subgenome preferences at different translation and PTMs levels (Supplementary Fig. S7D–G). While the light-induced transcriptome exhibited an overall C-dominance in differential expression, the proteome was preferentially A-dominated. Consistent with protein accumulation, PTM-A and PTM-U were predominantly associated with the A sub-genome, whereas PTM-P showed no significant subgenome preference (Supplementary Fig. S7I). This divergence underscores the role of extensive post-transcriptional regulation, which could partially explain the low correlation observed between the transcriptome and proteome (Pearson coefficient r ≈ 0.2).

Collectively, the observed patterns—temporal partitioning, family-specific co-expression, and subgenome bias—demonstrate that the homeologous gene in B. napus have evolved beyond simple redundancy toward sophisticated functional diversification. This subfunctionalization likely fine-tunes chloroplast development, ultimately orchestrating the crop’s enhanced photosynthetic traits.

These observations are consistent with functional studies in other polyploid crops, where homeologous genes guiding chloroplast-related traits exhibit complementary or partitioned roles. In hexaploid wheat, homeoallelic copies of TaCHLI encoding a chlorophyll biosynthesis enzyme function redundantly such that only double mutants show phenotypic defects [46]. Similarly, the chlorophyll-deficient phenotype in tetraploid tobacco requires mutation of both WS1A and WS1B homeologs [47]. Light receptor genes also show subfunctionalization in polyploids. soybean GmPHYA copies have diverged to control distinct facets of photomorphogenesis and flowering [48]. In B. napus, pan-genome analyses have revealed extensive structural variation between A- and C-subgenome homeologs [49], providing a genomic basis for the divergence observed here. Systematic functional interrogation of individual homeologous pairs using CRISPR-based approaches will be an important next step toward establishing causality.

Chlorophyll biosynthetic pathway temporally coordinates with chloroplast development

The transformation of proplastids into mature, photosynthetically active chloroplasts is not a linear process, but a finely tuned developmental program requiring precise spatiotemporal coordination of multiple biosynthetic pathways. To map this coordination, we profiled the abundance of 21 enzyme genes—involved in 15 steps of tetrapyrrole (chlorophyll and heme) biosynthesis—at the transcriptional, translational, and post-translational levels (Fig. 5 and Supplementary Figs S8 and S9). PTM dynamics function as a fast, early regulatory layer that precedes and correlates with subsequent protein accumulation, temporally orchestrating flux through the heme and chlorophyll branches of tetrapyrrole biosynthesis. PTM-P and PTM-U drive early temporal responses. PTM-A maintains basal or late-stage stability matching protein accumulation. We also considered genes involved in chloroplast biogenesis, including photoreceptors, light-signaling components, TFs, and plastid division factors [50]. Strikingly, global expression analysis revealed that chlorophyll biosynthesis genes did not form a separate cluster; instead, their accumulation patterns were intertwined with those of transcriptional regulators and structural genes mediating chloroplast development (Supplementary Fig. S8A and B).

Figure 5.

Multi-level sub-functionalization fine-tunes the chlorophyll biosynthetic pathway. Integrated temporal profiles of transcript (RNA), protein, and PTM levels for duplicated genes, illustrating cooperative regulation. (A) Schema of the chlorophyll biosynthesis pathway, highlighting enzymes and regulatory steps analyzed in this study. (B, C) Heatmaps illustrating the expression and modification dynamics of genes associated with the chlorophyll pathway (B) and the heme pathway (C) across the experimental time points. Left panel: Log₂-transformed transcript abundance (grayscale; darkest = maximum across all time points). Right panel: relative abundance of each gene/protein (yellow–purple scale; independently scaled 0–1 across time points), illustrating temporal accumulation patterns. White boxes indicate RNA, protein, or PTM with no detectable signal. For each protein with multiple modification sites, their values are summed prior scaling between 0 and 1.

Compared to the proteome, the transcriptome analysis yielded clear, stage-specific co-expression clusters. Most of the proteins were highest expressed at 6 and 12 h, while the expression profiles of PORA, PORB, PORC (protochlorophyllide oxidoreductase), and divinyl reductase (DVR) showed clear divergence (Supplementary Fig. S8A and B). These proteins participate in an essential light-dependent step catalyzed by POR and DVR (Fig. 5). While the three POR isoforms (PORA, PORB, and PORC) show evidence of subfunctionalization, their transcript accumulation profiles differ. Specifically, PORA transcript levels are relatively low compared to the other two isoforms (Fig. 5). PORA expression and protein accumulation are consistent, and it is co-expressed with early light-signaling components (e.g. CRY1, PIFs, PHYA). The PORB protein—along with PORA—is pre-loaded at the onset of illumination (0–2 h) (Supplementary Fig. S8B), providing immediate capacity for the photoconversion of existing protochlorophyllide. In contrast, PORC and DVR accumulate predominantly at later stages (6–12 h), driving the production of new enzyme to sustain chlorophyllide synthesis. The bimodal abundance peak of the DVR protein (at 0.5 and 12 h) suggests roles in both initial activation and maintenance of metabolic flux. A resource-efficient expression strategy for chlorophyll biosynthetic enzyme-encoding genes upon illumination involves either initially slight transcript accumulation or subfunctionalization within gene families. The transcript accumulation of PORB—the most abundant transcript showing signs of subfunctionalization—diverges from its protein accumulation pattern, a discrepancy explained in part by dynamic post-translational regulation. Analysis of the five expressed PORB genes revealed that the major isoform, A01.PORB, undergoes extensive modification: peak protein abundance at dark coincided with peak acetylation, phosphorylation peaked at half hour after illumination, and ubiquitination was lowest at that time (Fig. 5). In total, we mapped 11 acetylation, 8 phosphorylation, and 4 ubiquitination sites on A01.PORB (Supplementary Fig. S9), highlighting a complex, time-resolved PTM landscape. At the pathway level, protein acetylation was notably prevalent, implicated in most biosynthetic steps (10/13), whereas ubiquitination was restricted to only three proteins, including PORB(C07/A01) and CHLG. Together, these results reinforce a model in which acetylation generally enhances protein activity, while ubiquitination reduces protein stability.

Collectively, our data demonstrate that chlorophyll biosynthesis is a complex, light-regulated process. It involves the dynamic cooperation of genes, with an initial response followed by the staggered deployment of specific genes and partner enzymes, all fine-tuned by PTMs to efficiently manage the metabolic transition from dark to light.

An intricate, multi-layered regulation orchestrates the precise biogenesis and assembly of core photosynthetic complexes

Photosynthesis, the foundational process sustaining plant viability, relies on the precise assembly and spatiotemporal regulation of multi-protein complexes—including light-harvesting antenna complexes (Lhca/Lhcb), photosystem cores (PSI/PSII), and the cytochrome b6f (cytb6f) complex, which mediates electron transport. The functional integration of these membrane complexes necessitates harmonized action across multiple layers. Based on sequence homology with A. thaliana, we systematically identified the gene repertoire. We then profiled transcriptional, translational, and PTMs data to generate a comprehensive map of 61 PSII-, 50 PSI-, 39 antenna-, and 8 cytb6f-encoding genes (Fig. 6), including the chloroplast-encoded core subunits PsaA, PsaB, PsbA, PsbB, and PsbC. This analysis underscores the dual genomic contribution required for photosystem biogenesis and establishes an initial framework for understanding photosynthetic machinery assembly.

Figure 6.

Multi-level sub-functionalization fine-tunes the photosynthesis pathway. Joined temporal profiles of transcript (RNA), protein, and PTM levels for duplicated genes, illustrating cooperative regulation. (A) schema of the photosynthesis, highlighting enzymes analyzed in this study. (i) represents PSII, (ii) represents Cytb6f, (iii) represents PSI, and (iv) represents Antenna. (B) Heatmaps depicting the expression and modification dynamics of pathway-associated genes across experimental time points. Left panel: Log2-transformed transcript abundance (grayscale; darkest = maximum across all time points). Right panel: relative abundance of each gene/protein (yellow–purple scale; independently scaled 0–1 across time points), illustrating temporal accumulation patterns. White boxes indicate RNA, protein, or PTM with no detectable signal. For each protein with multiple modification sites, their values are summed prior scaling between 0 and 1.

In contrast to chlorophyll biosynthesis genes, which are transcriptionally correlated with chloroplast biogenesis, photosynthetic membrane proteins form a distinct accumulation cluster separate from chloroplast biogenesis proteins by 6–12 h—a time point marking the completion of chloroplast development (Supplementary Fig. S8D). This divergence is reflected transcriptionally, where only a small subset of the corresponding transcripts exhibits brief light induction (e.g. PSBC, PSBK, PSB27, PSBA), whereas PSBE is expressed in the dark (Supplementary Fig. S8C). Further analysis of transcript accumulation across gene families revealed temporal heterogeneity, partitioning genes into a minority induced only after short illumination and a majority requiring prolonged light exposure. Early-response transcripts peak transiently within 0–0.5 h but remain at low levels (e.g. A08.PSAO, C08.PSAO, C03.PSAB, C09/07.Lhcb6), while late-response transcripts initiate accumulation at 2 h, show steady increased expression, and support synchronized protein synthesis by 6 h. This conserved transcriptional program likely enables photosynthetic establishment through transcriptional regulation, with potential fine-tuning by transcript stability.

The major structural components of PSI and PSII—including reaction center subunits and light-harvesting proteins—all reached peak abundance within a narrow 6–12 h window at the protein level (Fig. 6). Interestingly, PTMs—particularly acetylation and phosphorylation—follow nearly identical late-stage accumulation trajectories (Fig. 6 and Supplementary Fig. S10), forming a highly correlated PTM landscape that modulates the activity and stability of assembled complexes. This pattern suggests a tightly calibrated proteostatic bottleneck in which precisely timed translation of stabilized transcripts enables efficient photosystem assembly. Phosphorylation of Lhca/Lhcb proteins (6/12), e.g. acts as a master rheostat, tuning energy transfer by adjusting inter-complex connectivity and antenna stoichiometry. Furthermore, we observed a striking temporal convergence: regardless of the specific gene or PTM type, the major sites of modification accumulated in a parallel manner, culminating between 6 and 12 h post-induction (Supplementary Fig. S10). Collectively, these findings suggest that acetylation and phosphorylation status acts as a robust checkpoint essential for efficient photosynthetic function.

Together, this paradigm prioritizes robustness, ensuring high-fidelity synthesis of photosynthetic machinery through cooperative nucleus–chloroplast gene expression, followed by PTM-mediated tuning. Such cross-tier regulation maintains photosynthetic integrity under light condition and underscores that transcriptional–dependent processes establish a necessary framework, which is then decisively refined through PTM.

Light-activated, multi-layer dynamic regulation optimizes the activity of the Calvin cycle

The Calvin cycle serves as the ultimate biochemical conduit through which captured solar energy drives the conversion of carbon dioxide into organic sugars. This process of carbon fixation creates the essential biomass that fuels the growth and development of B. napus. The cycle comprises 12 enzymatically catalyzed steps (Fig. 7). We focused on genes encoding Rubisco subunits—the central carboxylase—and the enzymes catalyzing each sequential reaction. We analyzed the transcriptional, translational, and PTM dynamics for these genes following light induction during chloroplast development (Fig. 7 and Supplementary Fig. S11).

Figure 7.

Multi-level sub-functionalization fine-tunes the Calvin cycle pathway. Joined temporal profiles of transcript (RNA), protein, and PTM levels for duplicated genes, illustrating cooperative regulation. (A) Schema of the Calvin cycle, highlighting enzymes and regulatory steps analyzed in this study. (B) Heatmaps depicting the expression and modification dynamics of pathway-associated genes across experimental time points. Left panel: Log2-transformed transcript abundance (grayscale; darkest = maximum across all time points). Right panel: relative abundance of each gene/protein (yellow–purple scale; independently scaled 0–1 across time points), illustrating temporal accumulation patterns. White boxes indicate RNA, protein, or PTM with no detectable signal. For each protein with multiple modification sites, their values are summed prior scaling between 0 and 1.

Transcriptome profiling revealed that six genes exhibited a consistent late-phase accumulation: RbcS1B (Rubisco small subunit), GapA/B (glyceraldehyde-3-phosphate dehydrogenase), FBA1 (fructose-bisphosphate aldolase), HCEF1 (hydroxymethylbilane synthase), EMB3119 (a conserved hypothetical protein linked to carbon metabolism), and SBPASE (sedoheptulose-bisphosphatase). Transcript accumulation for these loci was evident within 2 h post-illumination, increased gradually through 6 h, and peaked at 12 h (Fig. 7). Notably, their protein pattern preceded a delayed accumulation of the corresponding proteins, which became readily detectable at 6 h and increased through 12 h, indicating translational follow-through. In contrast, other Calvin cycle genes displayed heterogeneous expression—transient early induction, oscillatory patterns, or stable transcript levels—suggesting divergent regulatory inputs. Among all genes assayed, RbcS transcripts reached the highest absolute levels, consistent with Rubisco’s role as the primary bottleneck in carbon fixation [9]. The pronounced transcript accumulation of RbcS likely reflects a high biosynthetic demand to maximize carboxylation capacity during sustained photosynthesis. Collectively, our data delineate a phased transcriptional program that optimizes Calvin cycle output in B. napus. Early-responsive genes may prime the pathway, while late induction of catalytic and modifying components—particularly Rubisco—ensures enhanced flux under prolonged light.

Protein level beyond transcriptional control, PTM analysis revealed a complementary layer of regulation that modulates enzyme activity independently of protein abundance. We identified extensive light-dependent phosphorylation events on several core enzymes. Specifically, RbcS1B and the DRT gene product exhibited phosphorylation on both Ser and Thr residues. Additional central metabolic enzymes—including RPI2 (Ribose-5-phosphate isomerase 2), TPI (Triose phosphate isomerase), EMB3119, FBA3 (Fructose-bisphosphate aldolase 3), PGK2 (Phosphoglycerate kinase 2), TKL1(Tousled-like kinase 1), and PGK3 (Phosphoglycerate kinase 3)—were phosphorylated specifically on serine residues (Supplementary Fig. S11). Individual proteins frequently harbored multiple dynamic phosphorylation sites, with site-specific occupancy varying across light-exposure durations. Moreover, abundant acetylation and ubiquitination events were detected, implicating these PTMs in the fine-tuning of protein stability and function. The dynamic accumulation of these modifications following light induction underscores a complex, multi-tiered network underpinning Calvin cycle activity.

In the Calvin–Benson cycle pathways of B. napus, a clear temporal sequence exists between PTM dynamics and protein accumulation. Specifically, protein and transcript levels increase at a later stage (6–12 h), whereas phosphorylation and ubiquitination change rapidly during the early phase (0.5–2 h). In contrast, acetylation remains stable, serving as a basal state. This temporal pattern is consistently observed across multiple enzymes, including RbcS, PGK2, GapA/B, FBA, PRK, TKL, RPE, and SBPASE. Collectively, these findings indicate that PTMs function as early, rapid regulators that precede and coordinate delayed protein accumulation, thereby enabling rapid metabolic tuning followed by sustained pathway function.

Upstream regulators of chlorophyll and photosynthesis pathway genes

To capture regulators that orchestrate the expression of chlorophyll biosynthesis, photosystem core complex, and Calvin cycle genes, we constructed a comprehensive network. This approach incorporates transcriptional co-expression GRNs and PPIs data from all differentially accumulated molecules in multi-omics, addressing the complementary function and inconsistent accumulation of these molecules to more fully map the regulatory system. After removing redundant connections generated from different times point or different omics datasets, we construct a comprehensive interaction landscape. From this consolidated network, we systematically extracted all interactions directly associated with our core gene set, ultimately identifying a sophisticated regulatory cohort comprising 91 genes, including 42 TFs and 18 transcriptional regulators (Supplementary Table S3). This finding points to a complex, multi-layered modulatory system.

Top hubs were defined by the number of their photosynthetic target genes, with counts aggregated for all members within the same gene family to account for the polyploid genome and multi-homeolog structure of B. napus. Based on this criterion, the ten highest-ranked hubs were identified as Golden2-Like GLK1 (3 members), GLK2 (4), GATA Transcription Factor GATA21 (6), B-Box Domain Protein BBX15 (2), NAC Domain Containing Protein ANAC089 (2), HY5 (1), BnaA09G0715800ZS (an uncharacterized regulator of interest), CCCH-type zinc finger protein AtC3H33 (1), MYB Domain Protein MYB106 (1), and Nuclear Factor Y subunit NF-YA7 (1). These genes are predicted to have high potential for influencing photosynthetic efficiency and yield in B. napus. Additionally, we identified other regulators previously implicated as upstream factors in this pathway, including HYH (2), ABI4 (4), PIF4 (4), and Phytochrome Interacting Factor PIF8 (1) [13, 35]. Furthermore, CONSTANS-Like COL2 (2; homologous to the flowering-time gene CONSTANS (CO) [51] and TT8 (2; involved in seed color determination, where its knockout directly reduces oil production) [36] may have potential for cross-talk within this network.

To provide experimental support for the computationally inferred regulatory network, we performed transient dual-luciferase assays in B. napus mesophyll protoplasts. Reporter constructs containing the promoters of BnLHCB6, BnMDH1, and BnPETC driving firefly luciferase (LUC) were co-transformed with effector constructs expressing BnGLK1-YFP, BnMYB44-YFP, and BnBBX15-YFP, respectively. In all three cases, co-expression of the effector significantly enhanced reporter activity relative to the YFP empty vector control (Supplementary Fig. S12), confirming that these TFs can directly activate promoters of distinct photosynthesis-associated genes.

We further examined whether subgenome-divergent promoter activity contributes to the observed transcriptional bias between A and C subgenomes. LUC reporters driven by the A08 and C08 promoters of BnLHCB6 were co-transformed with BnGLK1-YFP or the YFP control. BnGLK1-YFP activated both promoters; however, activation of the A08 promoter was significantly stronger than that of the C08 promoter (Supplementary Fig. S12). This differential response is consistent with the subgenome-biased expression of BnLHCB6 homeologs observed in our transcriptomic data. Collectively, these assays demonstrate that cis-regulatory divergence between homeologous promoters contributes to subgenome-specific expression patterns during chloroplast development.

In conclusion, our integrated multi-omics network analysis has mapped a comprehensive set of regulators associated with the photosynthetic apparatus during chloroplast development in B. napus. It provides a powerful resource for future functional studies aimed at deciphering the precise mechanisms of photosynthetic gene expression and for engineering enhanced photosynthetic efficiency in crops.

Discussion

Chloroplast biogenesis and functional maturation are pivotal for plant autotrophic growth. Their regulation by light is critical for establishing photosynthetic efficiency and overall plant development [2]. While studies in model diploids like Arabidopsis and Maize have laid the groundwork for understanding chloroplast development during de-etiolation, their reliance on single-omics approaches has provided only fragmentary insights, thus yielding a necessarily partial view of any given layer [52–54]. As a result, our understanding of how the transcriptional, translational, and post-translational tiers are temporally associated and functionally organized during this dynamic process remains lacking. This gap is particularly pronounced in polyploid crops, where the regulatory complexity is amplified by genome duplication and subfunctionalization of homeologous genes [14, 15].In this study, we characterized the phenotypic responses of chloroplast development to light duration in tetraploid B. napus and elucidated a suite of multi-layered regulatory mechanisms through comprehensive multi-omics analyses. Our findings contribute to understanding of the interplay of transcriptional, post-transcriptional, translational, and post-translational processes to optimize chloroplast function during light-driven developmental transitions.

The orchestration of gene expression by the spatiotemporal dynamics of TFs accumulation enables cells to respond with precision to developmental and environmental signals [55]. Our findings align with this paradigm and futher indicate that the timing and sequence of TF activity schedule plastid maturation. We identified hundreds of TFs that dynamically accumulate in distinct stages, and importantly, this temporal progression dictates the order, duration, and intensity of downstream transcriptional events. This, in turn, confers specificity to the regulatory network that converts proplastids into functional chloroplasts. Unlike previous static analyses have cataloged individual TFs involved in this process [56–58], our time-resolved approach reveals the underlying temporal logic—specifically, how early-acting TFs establish a permissive landscape that licenses the subsequent activation of intermediate and late-stage factors. Within this framework, we have identified a previously undescribed network expansion event, wherein late-acting TFs are recruited to amplify and refine the initial transcriptional cues. This model reinforces the concept that complex cellular structures are built not through single regulatory switches but through stabilized sequences of TF activity that canalize developmental trajectories. Here, our work provides a mechanistic framework for understanding how external signals such as light are transduced through dynamic TF cascades to direct organelle differentiation.

The observed inconsistency between transcript abundance and protein accumulation, a well-documented phenomenon, stems not only from methodological variances but also from fundamental differences in the regulation and functional timing of mRNAs and their protein products [59–62]. Our data directly demonstrate that protein accumulation lags behind transcriptional induction across light-responsive modules, forming a temporal buffer. While light induces a rapid transcriptional wave (0–6 h), the photosynthetic program is translationally gated (2–12 h). This temporal decoupling—a core regulatory principle of chloroplast biogenesis—reflects the central dogma of molecular biology at a fundamental level. Meanwhile, we propose this phased execution acts as an active quality-control buffer, ensuring that the energetically costly assembly of photosynthetic complexes proceeds only under sustained favorable light, thereby optimizing plant fitness [63]. This lag provides an opportunity for post-transcriptional and translational mechanisms—such as AS, RNA decay, and translational modulation—to operate.

Additionally, light-induced differential accumulation revealed C-subgenome dominance at the transcriptome level but A-subgenome bias at the proteome level, indicating divergence between transcriptional and post-transcriptional regulation in polyploids. Although subgenome expression asymmetry is well documented at the RNA level in allopolyploids such as wheat, cotton, and common carp [64, 65], the reversal of dominance between omic layers suggests substantial post-transcriptional buffering. Several non-mutually exclusive mechanisms could explain this reversal. First, differential translational efficiency may allow lower-abundance A-subgenome transcripts to be translated more effectively—for instance, through favorable 5′ UTR structures or codon usage—while cis-regulatory dynamics contribute to transcriptional divergence [65, 66]. Supporting this, RNA structure modulates translational asymmetry in tetraploid wheat, where single-nucleotide variations between homoeologs directly affect translation output [67]; similar mechanisms may operate in our system. Second, differential mRNA stability could also play a role: C-subgenome transcripts might be degraded faster despite higher synthesis rates [65], meaning that RNA-level processing shapes the pool of translatable transcripts. Additionally, transposable elements and small RNA-mediated silencing could contribute to transcriptional subgenome dominance. Finally, differential protein degradation could reinforce A-subgenome bias: reduced ubiquitination of A-subgenome homoeologs may prolong their protein half-lives, leading to higher steady-state protein levels even when synthesis rates are comparatively low. We propose that future work incorporating ribosome profiling, mRNA decay assays, and pulse-chase proteomics will be essential to dissect the regulatory layers steering subgenome expression dynamics—a critical step toward understanding how polyploid genomes align gene expression across multiple levels to achieve coherent phenotypic outcomes.

PTMs, including acetylation, phosphorylation, and ubiquitination, represent an important post-translational layer for chloroplast development by directly modulating protein stability and activity [68–71]. While historically studied for discrete roles in photosynthetic acclimation [52, 72], our findings position PTMs within a broader system that spans cellular compartments. In B. napus, PTMs function synergistically with transcriptional and translational processes to form a robust, cross-tier regulation. This regulation exhibits functional compartmentalization; for instance, nuclear-localized phosphorylation is enriched in pathways linked to AS, which dynamically influence gene transcripts for photoreception, chlorophyll synthesis, and photosynthetic complex assembly. This positions kinase-mediated signaling as a modulator of nuclear post-transcriptional mechanisms that influence cellular activities essential for chloroplast biogenesis. It offers a new insight on nuclear chloroplast signal cross talk. This rapid, reversible nature of PTMs—accumulated independently of protein abundance—provides a precise mechanism for the precise coordination required during chloroplast biogenesis.

The tetraploid genome introduce both complexity and functional redundancy to its systems [73]. The presence of numerous homeologous gene pairs, combined with sub-genome expression bias and subfunctionalization, diversifies the genetic toolkit available for B. napus chloroplast development beyond that of diploid models like Arabidopsis, while simultaneously providing genetic buffering. This is exemplified by three biological processes: the temporal synchronization of chlorophyll biosynthesis with organelle development, the multi-layered orchestration of photosynthetic complex assembly, and the light-activated fine-tuning of Calvin cycle activity. Furthermore, our comprehensive analysis—combining GRNs and PPIs data—identified upstream regulators of these core pathways. These regulators likely function through complementary interactions across transcriptional and post-translational tiers. Together, these findings illustrate how B. napus incorporates its genomic complexity with dynamic PTM, transcriptional, and post-transcriptional regulation associated with the efficient biogenesis and function of its photosynthetic apparatus, offering insights into the evolution of robust networks in polyploids.

The 12 h window examined here primarily resolves the early signaling and structural establishment phase of chloroplast biogenesis. Thylakoid assembly and full photosynthetic maturation are known to extend over several days [74], with a subsequent chloroplast proliferation phase involving continued membrane expansion and organelle multiplication [4, 75]. Extended time courses would be valuable to capture the transition to proteostatic homeostasis and secondary remodeling of regulatory networks [2]. The high-resolution atlas presented here provides a foundation for such future studies.

In summary, this study generates a detailed molecular map of chloroplast development in a polyploid crop. It elucidates a sophisticated, multi-layered paradigm in rapeseed, integrating temporally decoupled transcriptional cascades, crosstalk between AS and PTMs, and homeologous gene subfunctionalization to ensure robust chloroplast biogenesis. A sequential TF cascade establishes the core temporal framework, while novel PTM–AS interactions fine-tune gene expression. Furthermore, PTMs contribute to protein stability involved in pathways during chloroplast development. Crucially, the subfunctionalization of homeologous genes reveals a polyploid-specific strategy for plasticity, providing a mechanistic basis for enhanced environmental adaptability. Regulatory nodes—especially upstream regulators of photosystem pathways—represent prime targets for enhancing photosynthetic efficiency and stability under variable light, an important trait for climate resilience. Collectively, these findings advance our understanding of polyploid plant biology and offer a multipurpose resource for future research. This includes mechanistic studies of specific pathways—such as those previously characterized for phytochromes and pigment synthesis—as well as comparative omics investigations [76].

Supplementary data

Supplementary data is available at NAR online.

Conflict of interest

None declared.

Funding

This work was supported by the Key Research and Development Program of Zhejiang Province (2024SSYS0100 to R.L.), the National Natural Science Foundation of China (U25A20633 to R.L. and 32570283 to Y.J.), and the Leading Innovation and Entrepreneurship Team Project of Hangzhou City (TD2024008 to R.L.). Funding to pay the Open Access publication charges for this article was provided by the National Natural Science Foundation of China.

Data availability

The RNA raw sequence data reported in this paper have been deposited in the Genome Sequence Archive in National Genomics Data Center [77, 78],China National Center for Bioinformation/Beijing Institute of Genomics, Chinese Academy of Sciences (GSA: CRA037016) that are publicly accessible at https://ngdc.cncb.ac.cn/gsa. The mass spectrometry proteome, acetyl-proteome, phospho-proteome, and ubiquitin-proteome data have been deposited to ProteomeXchange Consortium via the Integrated Proteome Resources (iProx) partner repository [79, 80] under the access IDs IPX0015290001, IPX 0015290002, IPX0015290003, and IPX0015290004, respectively.