Employing a hidden Markov model (HMM) profile corresponding to the Dof zinc-finger domain (PF02701), a HMMER3 search was performed on the COL-tetra database to identify potential CoDof candidate genes. The basic information of 40 CoDof family members, including 116 different gene isoforms. The alleles of all family members are paired and distributed between homologous chromosomes, and assign a CoDof ID to each gene based on chromosomal information and polyploidy homology. The number of amino acids of CoDof ranges from 84 (CoDof7.2) to 373 (CoDof17.2/CoDof18.2/CoDof21.2), with corresponding molecular weights varying from 9251.21 (CoDof7.2) to 41652.62 (CoDof18.2). The theoretical pI values show a wide distribution, from 4.72 (CoDof22.3) to 9.64 (CoDof13.2), indicating different acid-base properties. The instability index, aliphatic index, and grand average of hydropathicity are also provided. The instability index ranges from 26.58 (CoDof28.3) to 73.96 (CoDof17.3), reflecting the stability of the proteins. The aliphatic index varies between 33.08 (CoDof37/CoDof38.4/CoDof38.5) and 129.88 (CoDof7.2), and the grand average of hydropathicity ranges from -1.216 (CoDof17.1) to 0.351(CoDof7.2), which are related to hydrophobicity and structural characteristics of the protein. It is showed that except for CoDof7.2, the remaining 39 DoDof proteins all have a total average hydrophobicity index less than 1, indicating that most members of the Dof family are hydrophilic proteins. In addition, the physical and chemical properties of each allele are very similar, indicating that the functional genes in COL-tetra are relatively conserved ( Supplementary Table S1 ).

The MEME web-based platform was used to analysis 40 CoDof proteins with a total of 116 allele sequences uncovered three distinct conserved motifs. Three motifs displayed lengths ranging from 21 to 35 amino acid residues, and the variation in motif composition (number and type) across different CoDof proteins suggests functional diversification in vivo, possibly due to their distinct biological roles. Protein sequence alignment results show that the amino acid sequences in the characteristic domain regions of CoDof proteins exhibit relatively strong conservation, with three distinct forms: motif1 (GGCRK, indicated in red), motif2 (CPRCD, indicated in blue) and motif3 (indicated in green). In the characteristic domain of Dof proteins, a total of 28 amino acids are the most conserved, which may play a key role in contributing to the conservation of the Dof domain. Within the characteristic region of Dof proteins, there is one conserved motif (motif1) and one highly conserved C₂-C₂ type zinc finger structure (CX2CX21CX2C, Zinc-finger). The C₂-C₂ type zinc finger structure is located within the conserved motif (motif2). The four conserved Cys residues are an important feature of the zinc finger structure and a key characteristic for identifying the Dof domain. Furthermore, the conserved motif analysis of CoDof proteins shows that the amino acid sequences of all members contain motif1, which is located within the characteristic domain of the CoDof family, implying the conservation of the characteristic Dof domain among family members. Conserved motifs in CoDof members of the same equipotential exhibit a certain degree of similarity in composition and arrangement. However, it is particularly noted that CoDof38.1, CoDof38.4, and CoDof38.6 lack the conserved motif 2, whereas CoDof38.3 and CoDof38.5 possess the conserved motif 2 structure ( Figure 1 ).

Gene structure analysis revealed that the number of introns of the CoDof family members is between 0 and 2, There are 62 CoCofs have no introns, 52 CoDofs contain one intron, and only 2 genes contain two introns, namely CoDof20 and CoDof36.4. Besides, the lengths of UTRs and CDSs vary significantly among different members. For example, CoDof28.3 has a relatively long CDS compared to others. Some members, like CoDof1.1, CoDof1.2 and CoDof1.3, show different patterns of UTR and CDS lengths, indicating possible alternative splicing events. Different types of allele representation provide a comprehensive view of the structural diversity within the CoDof gene family ( Figure 2 ).

Phylogenetic tree cluster analysis of CoDof family members was performed using Arabidopsis thaliana and selected soybean GmDof family members as references. Based on Arabidopsis classification criteria, the Dof proteins was classified into 9 subfamilies, including A, B1, B2, C2, C2.1, C2.2, C3, D1, and D2 ( Figure 3 ). Among them, the A subclass contained the highest number of CoDof family proteins (8 members), followed by the B1 subfamily with 6 DoDof members. Genes within the family exhibit sequence homology and functional similarity across plant species. GmDof4 and GmDof11, associated with fatty acid and lipid biosynthesis, were located in the A and C2 subfamilies respectively, indicating that proteins in these two subclades are linked to lipid metabolic processes. Furthermore, significant differences in distribution proportions of family members exist among different species within the same subfamily, reflecting functional divergence of Dof proteins across taxa ( Figure 3 ).

The 2-kilobase promoter regions upstream of CoDofs genes and the actin housekeeping gene were computationally analyzed using the PlantCARE database to identify potential cis-acting regulatory elements ( Figure 4 ). This analysis revealed five categories of cell development-related cis-elements in CoDofs promoters, including endosperm expression, palisade mesophyll cells, flavonoid biosynthetic genes regulation, and cell cycle regulation motifs. Additionally, four hormone-responsive regulatory modules were identified, consisting of abscisic acid-, auxin-, and gibberellin -responsive elements, along with zein metabolism regulatory sequences. Notably, 20 stress-responsive cis-elements were characterized, encompassing light responsive elements, anaerobic induction, circadian control, anoxic specific inducibility, Ethylene-responsiveelement, elicitor-mediated activation, low-temperature responsiveness, drought-inducibility, defense and stress responsiveness, and wound-responsiveness ( Supplementary Table 2 ). Conservation analysis of the actin promoter region identified core regulatory elements such as ABRE, ARE, ERE, G-box, MYC, and STRE, highlighting the conserved regulatory features across COL-tetra genes.

To study the expression pattern and pathway of CoDof genes in COL-tetra, we first sampled the seed tissue of COL-tetra and determined the fatty acid content. The phenotypic characteristics seeds showing the fruits as ellipsoidal with a smooth surface, transitioning in color from green to red ( Figure 5A ). The average diameter of the seeds is approximately 19.12 mm ( Figure 5B ), with the kernel being dark brown and hard in texture ( Figure 5C ). The physiological changes during fruit development illustrates the changes in dry and fresh weight over time, with dry weight increasing rapidly in the early stages, peaking at 34 weeks, and then slightly decreasing ( Figure 5D ). Fresh weight remains relatively stable throughout the development period. The moisture content of the seed kernel, which decreases steadily from the initial stages ( Figure 5E ). The oil content of the seed kernel, which increases gradually, reaching its highest level at 51 weeks ( Figure 5F ). The changes in fatty acid composition, with oleic and linoleic acid contents increasing over time, while palmitic and stearic acid contents decrease ( Figure 5G ).

Besides, transcriptome sequencing was performed using RNA-seq based on different developmental stages of seeds. According to the Gene Ontology (GO) enrichment results, the obtained differential expression genes (DEGs) were significant enriched in biological processes such as fatty acid beta-oxidation, unsaturated even-numbered fatty acid metabolic process, raffinose catabolic process, and regulation of auxin biosynthetic process. These results indicate that fatty acid metabolism and carbohydrate transport are crucial biological processes during fruit development ( Figure 5H ). The Kyoto Encyclopedia of Genes and Genomes (KEGG) revealed that the DEGs were significant enrichment in pathways related to signaling and cellular processes, biosynthesis of other secondary metabolites, and amino acid metabolism. These findings suggest that signal transduction and metabolic pathways play important roles in the development of COL-tetra seeds ( Figure 5I ).

The synthesis of triacylglycerols (TAG) in COL-tetra involves intricate enzymatic systems and diverse chemical reactions, primarily mediated through the acyl-CoA-dependent Kennedy pathway and the acyl-CoA-independent phospholipid: diacylglycerol acyltransferase (PDAT) pathway. A total of 13 GPAT, 1 ATS, 3 PAP, 3 PAH, 9 LPAT, 1 DGAT1, 2 DGAT2, and 7 PDAT genes were identified during TAG biosynthesis ( Figure 6A ). Analysis of their expression patterns across seed developmental stages revealed three key insights: first, not all members of the same gene family are transcriptionally active, as exemplified by only 6 of the 13 GPAT genes being expressed, suggesting selective silencing of certain family members during lipid synthesis; second, dynamic temporal regulation governs gene expression, with PDAT1 exhibiting a fluctuating rise-fall-rise-fall pattern and PDAT2a displaying a simpler rise-fall trend, implying coordinated functional specialization within gene families; third, DGAT1, the terminal rate-limiting enzyme in TAG assembly, showed significantly higher expression than DGAT2, while PDAT pathway genes outnumbered and surpassed DGAT genes in both quantity and expression levels. This dual-pathway collaboration between PDAT and DGAT enzymes likely underpins the compositional diversity of COL-tetra lipids ( Figure 6A ).

Besides, the expression analysis of genes related to fatty acid biosynthesis in COL-tetra. For genes with multiple alleles, the expression levels of different alleles were summed. In the de novo fatty acid biosynthesis pathway, ACCase acts as a key rate-limiting enzyme. Multiple gene families encoding different subunits of ACCase were identified in COL-tetra. Expression profiling of these gene families during different seed developmental stages revealed that not all members of the same gene family were expressed - some family members remained silent during seed development. Additionally, expressed genes showed higher expression levels in early and late developmental stages compared to the third stage. However, seed fatty acid content continued to increase throughout development, with lower accumulation in early stages, suggesting that fatty acids synthesized in early stages were not fully converted into storage lipids. Multiple gene families involved in fatty acid chain elongation were also identified. Expressed genes within these families displayed similar expression patterns - higher in early and late stages, lower in mid-stage. Given the high oleic acid content (up to 80%) in COL-tetra oil, enzymes related to 18-carbon fatty acid desaturation were specifically analyzed. Nine SAD genes, 11 FAD2/FAD6 genes, and 16 FAD3/FAD8 genes were identified. Interestingly, despite oleic acid accounting for over 80% of total fatty acids (compared to ~8% linoleic acid and 0.5% linolenic acid), the gene family encoding stearoyl-ACP desaturase (SAD) showed the fewest members, while ω-3 fatty acid desaturase genes (FAD3/FAD8) were most abundant. This further confirms that gene copy number does not directly correlate with phenotypic traits. Expression analysis revealed significantly higher SAD gene expression across all developmental stages compared to FAD2/FAD6, which in turn showed higher expression than FAD3/FAD8. This expression hierarchy explains the characteristic high-oleic, low-linoleic, low-linolenic acid composition of COL-tetra oil ( Figure 6B ).

To investigate the functional diversity of CoDof gene family members throughout the developmental cycle of COL-tetra, we conducted expression pattern analysis using RNA-seq BAM files from multiple organ systems and developmental phases. The heatmap visualization revealed distinct spatial-temporal expression profiles among CoDof members, enabling their classification into eight expression clusters ( Figure 7A ). Notably, three genes (CoDof37, CoDof28, and CoDof35) exhibited preferential expression in root tissues, while CoDof11 demonstrated floral-specific activation. Two distinct expression patterns emerged in photosynthetic tissues, CoDof18, CoDof23, CoDof32, CoDof25, and CoDof7 showed meristem-associated expression in tender leaves, whereas CoDof9, CoDof39, CoDof14, CoDof29, CoDof40, and CoDof38 displayed dual activation in both tender leaves and leaf buds. Besides, seed developmental analysis uncovered stage-specific regulation: Five genes (CoDof4, CoDof6, CoDof16, CoDof33, and CoDof38) displayed late maturation-phase activation (16-week seeds), contrasting with four genes (CoDof10, CoDof22, CoDof30, and CoDof10) showing early developmental preference. Specialized expression patterns were observed for CoDof35 and CoDof26 in seed coat formation, while the remaining family members exhibited stem-enriched expression profiles. These differential expression patterns across developmental stages and tissue types suggest functional diversification of CoDof transcription factors in regulating growth and organogenesis processes in COL-tetra. The stage-specific activation of particular gene subsets implies specialized roles in temporal developmental regulation, while tissue-preferential expression indicates spatial functional differentiation within plant architecture.

Because on fatty acid/lipid pathways, Pearson correlation analysis was performed between fatty acid/lipid biosynthesis genes and CoDofs family members. The result revealed that CoDof11.1, CoDof40.1 and CoDof25.1 were strongly positively linked with fatty acid desaturase pathway, with correlation coefficient is greater than 0.9. The CoDof25.1, CoDof33 and CoDof7.1 were strongly positively linked with fatty acid biosynthesis initiation pathway, CoDof30.1 and CoDof22.1 was highly positively associated with long-chain acyl-CoA synthetase. Besides, the CoDof30.1, CoDof6.2, and CoDof4 was highly positively associated with triacylglycerol biosynthesis, while CoDof22.1 was strongly positively linked with fatty acid biosynthesis elongation pathway. Importantly, the CoDof29, CoDof38, CoDof40, and CoDof30 as candidate transcription factors highly positively associated with DGAT and SAD, which play crucial roles in de novo biosynthesis of fatty acids and lipid ( Figure 7B ).

To validate the expression patterns in seed differential stages of four candidate CoDof transcription factors (CoDof29.1, CoDof38.1, CoDof40.2, and CoDof30.1), quantitative reverse transcription PCR (RT-qPCR) was performed. Comparative analysis revealed strong concordance between the RT-qPCR results and the RNA-seq data, confirming consistent expression trends across both methodologies. Among them, CoDof29.1 and CoDof30.1 were highly expressed in the middle stage of seed development (44 weeks seed), while CoDof38.1 and CoDof40.2 were specifically expression in the early stage of seed development (36 and 41 weeks seed). This indicated that CoDofs may exert important regulatory effects in the fatty acids and lipid pathway during the stages of seed development in COL-tetra ( Figure 8A ). To investigate the transcriptional activation potential of CoDof30.1, we conducted yeast hybrid assays using Y2H Gold strains carrying BD-CoDof30.1 fusion constructs. Growth patterns comparable to positive controls were observed on selective SD/-Trp/-Ade/-His/-Leu media, demonstrating significant transcriptional activation capacity ( Figure 8B ). Subcellular distribution analysis was performed through transient expression of 35S::CoDof30-eGFP constructs in tobacco leaf epidermal cells. Confocal microscopy revealed distinct localization patterns: while control 35S::eGFP constructs showed ubiquitous cytoplasmic and nuclear fluorescence, and the CoDof30 fusion protein exhibited nuclear and membrane accumulation ( Figure 8C ). These findings confirm nuclear and membrane localization and functional role as a transcription activator of CoDof30, which considered as a key candidate Dof transcription factor involved in fatty acid and lipid process.

To elucidate the DNA binding characteristics of the one finger family (Dof) of transcription factors, the hidden Markov model (HMM) profile of the Dof domain, a zinc finger DNA-binding domain (PF02701), was utilized, sourced from the Pfam database (El-Gebali et al., 2019). The reference genomes of the COL-tetra served as the foundational basis for this investigation (Zhang et al., 2024). The presence of conserved domains within the Dof family members was validated using HMMER SEARCH 3.0 with a cutoff E-value of ≤ 0.01, in conjunction with the SMART database (Finn et al., 2011a). Theoretical isoelectric point (pI) and molecular weight (Mw) values were computed using the ExPaSy online tool (https://web.expasy.org/compute_pi/) (Finn et al., 2011b). For protein modeling, the SWISS-MODEL online tool (https://swissmodel.expasy.org/) was employed (Gasteiger et al., 2003), while the putative subcellular localization of genes was predicted using the Softberry online software (http://www.softberry.com).

Conserved motifs were identified utilizing the MEME online tool (Version 5.1.0, National Institutes of Health, Bethesda, MD, USA), and the aligned domains were visualized with DNAMAN (Version 8.0.8, Lynnon Biosoft, San Ramon, CA, USA). The gene structures of CoDofs were predicted using the Gene Structure Display Server (Kumar et al., 2018). An unrooted phylogenetic tree for CoDofs and AtDofs proteins was constructed using Maximum Likelihood (ML) method, with bootstrap testing conducted over 1000 replicates, using MEGA11 (https://www.megasoftware.net/home) (Chen et al., 2023).

The promoter regions, defined as the 2000 base pair sequences upstream of the coding DNA sequences (CDS), were examined for cis-regulatory elements utilizing the PlantCARE web tool (http://bioinformatics.psb.ugent.be/webtools/plantcare/html/) (Lescot et al., 2002).

The seed materials for this research were sourced from 8-year-old Camellia oleifera trees planted in the experimental zone of Central South University of Forestry and Technology, located in Jiayi Town, Pingjiang County, Hunan Province, China (113°51’1” N, 28°38’21’’ E). This area experiences a subtropical monsoon climate with an annual frost-free period of approximately 275 days. The average yearly temperature is 16.8°C, and the average annual rainfall is 1450.8 mm. Seed samples were gathered at five different developmental periods: 36, 41, 44, 48, and 51 weeks after flowering (WAF), and were stored at -80°C until needed. The raw RNA-seq data (GWHERBT00000000) used in the study were sourced from the National Genomics Data Center (https://ngdc.cncb.ac.cn/) (Zhang et al., 2024). GO enrichment analyses were conducted using the Plant Transcriptional Regulatory Map (Jin et al., 2017), with a significance level set at a corrected p-value < 0.05. KEGG analyses were performed using KOBAS-i: intelligence online tools (http://bioinfo.org/kobaskobas3/?t=1) and visualized with R (Bu et al., 2021).

C. oleifera seeds were oven-dried at 60°C to a constant weight, ground into powder, and 0.1 g was placed in a 15 mL centrifuge tube. After adding 4 mL n-heptane and 100 μL of a C12:0 internal standard, the mixture was vortexed for 15 seconds and rested for 30 seconds, repeating this cycle three times. Then, 200 μL of 2 mol/L potassium hydroxide methanol solution was added and vortexed until clear. After adding 1 g anhydrous sodium sulfate and vortexing again, the supernatant was filtered through a 0.45 μm filter and prepared for gas chromatography analysis.

The gas chromatography (GC) analysis was conducted utilizing a Shimadzu GC2014 system (Kyoto, Japan), which was equipped with a DB-WAX column (30 m; Agilent Technologies, Palo Alto, CA, USA). Nitrogen was employed as the carrier gas at a split ratio of 20:1. Detection was achieved using a hydrogen flame ionization detector, maintained at a temperature of 230°C. The column temperature program was configured as follows: an initial temperature of 100°C was maintained for 5 minutes, followed by an increase to 220°C at a rate of 10°C per minute, with a final hold at 220°C for 15 minutes. The oil content was calculated using the formula: Oil content = [(4/percentage of internal standard (C12:0) - 4)/weight of powder] × 100%. For the determination of oil content, three biological replicates were performed for each developmental period.

Total RNA was extracted using the TransZol kit (TransGen Biotech, Inc., Beijing, China), and complementary DNA (cDNA) synthesis was conducted with the HiScript II QRT SuperMix for qPCR (+gDNA wiper) kit from Vazyme (Piscataway, NJ, USA). DNase was used DNase I, RNase-free from Vazyme (Piscataway, NJ, USA). RT-qPCR experiments were performed utilizing the LightCycler 96 Real-Time PCR System from Roche (Basel, Switzerland). A reaction volume of 50 μL was prepared, with each sample subjected to three biological replicates and three technical replicates. The HiScript II Q RT SuperMix for qPCR (+gDNA wiper) kit from Vazyme was used for reverse transcription. Relative expression levels were calculated using the 2^−ΔΔCt method, with CoEF1α serving as the internal reference gene for normalization. Expression levels were normalized to the sample exhibiting the lowest expression. Details of the primers for CoDofs and the CoEF1α gene are provided in Supplementary Table S3 . Data are presented as mean ± standard deviations (SD), with statistical significance determined by Student’s t-test (* p < 0.05, ** p < 0.01, *** p < 0.001).

The Pearson correlation coefficient (r) was employed to assess the correlation between fatty acids and lipid biosynthesis genes and CoDof transcription factors. A higher absolute value of r indicates a stronger correlation, with values exceeding 0.5 depicted in Figure 6A . The complete coding sequence (CDS) of CoDof30.1 was cloned into the pGBDKT7 vector. Subsequently, the constructs pGBDKT7- CoDof30.1, pGBDKT7-53 (positive control), and pGBDKT7-lam (negative control) were introduced into the Y2HGold yeast strain. The transformed yeast cells were cultured on three types of media: SD/-Trp and SD/-Trp/-His/-Ade plates. Details of the primers used are provided in Supplementary Table S3 .

The complete coding sequence (CDS) of CoDof30.1, excluding the stop codon, was inserted into the pSUPER1300 vector and subsequently fused with the green fluorescent protein (GFP) to create a CoDof30.1-GFP fusion protein. The empty pSUPER1300 vector was utilized as the control. These constructions were introduced into Agrobacterium tumefaciens strain GV3101. Co-transformation of the GFP-fusion constructs and the nuclear localization marker (NF-YA4-mCherry) was conducted in Nicotiana benthamiana leaves. After 48 hours post-infiltration, fluorescence signals were detected at excitation/emission wavelengths of 488/510 nm for GFP and 552/610 nm for mCherry using a confocal laser scanning microscope (Leica SP8, Heidelberg, Germany). Primer details are provided in Supplementary Table S3 .