Integrative metabolome and transcriptome of different tree size reveal wood formation in Corymbia citriodora

Abstract

Background

Corymbia citriodora wood holds extensive application value; however, due to the cross-pollination nature of C. citriodora and the significant genetic variation among offspring, a phenomenon is observed in plantations where trees with the same parental lineage exhibit varying diameter at breast height (DBH). In this study, we hypothesized that the observed DBH variation reflects two divergent adaptation strategies under winter conditions: smaller trees may exhibit heightened cold sensitivity, activating stress responses and entering early dormancy, whereas larger trees may sustain cambial activity through coordinated regulation of cell-cycle genes and carbohydrate metabolism.

Results

Metabolomics showed large trees enriched in phenolic compounds like picroside I and orientin. Transcriptomics revealed their genes were enriched in cell cycle and phenylalanine/tyrosine metabolism pathways, whereas small trees were enriched in abscisic acid (ABA) pathways. Integrated analysis revealed three patterns of transcript-metabolite coordination: (i) concordant upregulation (e.g., shikimate dehydrogenase (SDH)-shikimic acid), (ii) discordant regulation suggesting post-transcriptional control (raffinose accumulation without raffinose synthase (RFS) upregulation in small trees), and (iii) pathway-level flux redirection (phenylpropanoid pathway). Systematically, during winter, small trees exhibited stress-sensitive signatures, characterized by 9-cis-epoxycarotenoid dioxygenase (NCED), pyrabactin resistance/pyrabactin resistance-like (PYR/PYL), ABA-responsive element binding factor (ABF) upregulation, accumulation of raffinose, and transcriptional shift toward lignification (cinnamyl-alcohol dehydrogenase (CAD), peroxidase upregulation). Conversely, large trees showed molecular profiles indicative of stronger adaptability, with transcriptional signatures consistent with continued cell division (E2F transcription factor 3 (E2F3), mini-chromosome maintenance (MCM2/4/5/6) upregulation), cytoskeletal organization (kinesin KIN12B), and active carbohydrate metabolism (sucrose synthase (SUS), trehalose-6-phosphate phosphatase (TPP), alpha-amylase (AMY) upregulation), accumulating cinnamate and shikimic acid.

Conclusion

This study advances understanding of tree growth variation by demonstrating that winter conditions reveal divergent adaptation strategies linked to DBH. The key conceptual advance is the integration of stress signaling, carbon metabolism, and cell-wall regulation into a unified framework where differential dormancy timing explains within-family growth variation. Large trees maintain a permissive growth state through coordinated upregulation of cell-cycle, cytoskeletal, and biosynthetic genes, while small trees exhibit coordinated stress responses that accelerate dormancy. These findings provide species-specific targets for functional validation—including E2F3, KIN12B, and NCED—and suggest that breeding for sustained winter cambial activity could enhance biomass production in subtropical plantations.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12864-026-12886-6.

Background

Wood diameter at breast height (DBH) is a paramount economic trait in forestry, and breeders consider it a crucial criterion for evaluating superior trees. Nevertheless, a phenomenon exists where individuals within the same family exhibit varying DBHs. The differences in stem size can be attributed to interactions between genotype and environment [1, 2]. Across different forest tree species, the familial heritability of DBH is relatively low, for example, Populus euramericana has a heritability of 0.3 [3], and Castanopsis hystrix has a heritability of 0.38 [1], indicating moderate genetic control. Moreover, the substantial DBH variation among individuals within the same family can serve as a rich genetic resource. Compared to other traits, the variation in DBH may be governed by a larger number of genes.

In the process of forest tree growth and development, wood production undergoes primary and secondary growth, with secondary growth encompassing cambial differentiation and xylem formation. The increase in tree DBH, also known as radial growth, is accomplished by the activity of meristematic tissues in the vascular cambium [4]. Consequently, the vascular cambium can be used as a research material to uncover the potential mechanisms underlying radial growth. With the rapid advancement of molecular biology, the regulatory molecular mechanisms and associated genes of the cambium in many plants have been elucidated. For instance, ethylene synthesis-related genes in Arabidopsis thaliana affect vascular cambium development [5], disease resistance-related genes in Ginkgo biloba are highly expressed in the cambium of older trees compared to young trees [6], and overexpression of GRAS transcription factors promotes vascular cambium development in Arabidopsis [7]. Obvious differences in the structure and composition of the vascular cambium exist among trees with varying DBH. It has been reported that Ginkgo and Populus trees with smaller diameters have more cambial layers [6, 8], whereas in some species like Endospermum diadenum, trees with larger stem diameters exhibit longer growth periods and more cell layers than those with smaller diameters [9]. Additionally, in forest trees, DBH is significantly correlated with wood properties such as cellulose, lignin, and fiber traits [10, 11], indicating that research on radial growth can indirectly improve wood quality. As a model woody plant, Populus has revealed numerous genes associated with vascular cambium development, such as ERF85 [4], PttCLE47 [12], PtrHB4 [13], VCM1, and VCM2 [14], among others. These genes serve as references for other forest tree species; however, the differential genes related to cambium development in trees with different DBH still need to be identified.

The cells of the cambium have the capacity to divide, forming the phloem region outwardly and the xylem region inwardly. The xylem region undergoes processes such as cell division, secondary cell wall formation, lignification, and programmed cell death from the outside in [15]. In different wood-forming regions, there are significant differences in metabolite content. For example, in Populus, the levels of sucrose, phenylalanine, salicin, zeatin, and auxin are highest in the cambial region [16]. Sucrose plays a crucial role in the cambial region, as it has been reported to inhibit the differentiation of cambial cells into xylem and phloem [17]. Under low nitrogen conditions, plant growth rates significantly slow down, and studies have found that Populus trees under low nitrogen treatment exhibit increased sucrose and starch content in the cambium [18]. This suggests that under environmental stress, plants adjust their nutrient allocation strategies, thereby affecting wood formation. However, the differences in metabolites within the cambium of trees with varying DBH remain unknown.

Corymbia citriodora, a significant eucalypt species, plays a vital economic and ecological role in the southern regions of China. Its wood, characterized by straight grain and ease of processing, is highly favored in the production of medium-density fiberboard and hardboard [19, 20]. Additionally, the longer fiber length and higher pulp yield of C. citriodora make it an ideal raw material for the paper industry. Studies have indicated that the estimated heritability of DBH traits in C. citriodora is moderate, suggesting that these traits are influenced not only by genetic factors but also significantly by environmental factors [21]. A study conducted through association analysis has identified single nucleotide polymorphism (SNP) loci associated with DBH, which are linked to the SAM demethyltransferase (AtACL5) gene and the COP1-interacting protein 7 (AtCIP7) gene, both of which play key roles in xylem formation [22]. Currently, research on the metabolites of C. citriodora is primarily focused on leaves [23, 24], while the regulatory mechanisms of genes related to the cambium in wood development remain unexplored.

As a fast-growing tree with well-developed secondary vascular tissues, C. citriodora is an ideal material for studying wood formation. However, research on the differential metabolites and genes in wood of varying sizes in C. citriodora is still in its infancy (Fig. 1). In this study, we hypothesized that the observed DBH variation among genetically related C. citriodora individuals reflects two divergent adaptation strategies under winter conditions: smaller trees may exhibit heightened cold sensitivity, activating stress responses and entering early dormancy, whereas larger trees may sustain cambial activity through coordinated regulation of cell-cycle genes and carbohydrate metabolism. We further hypothesized that these differences would be reflected in distinct transcriptional and metabolic signatures, particularly in pathways governing phytohormone signaling, lignin biosynthesis, and carbon partitioning. In light of this, the present study selected the cambium of C. citriodora as the research subject to uncover the transcriptional regulation and molecular mechanisms underlying wood formation during winter. This research is expected to identify key genes influencing the rapid growth of eucalypts, providing new scientific insights for enhancing the biological properties and directed cultivation of wood in plantation forests.

Fig. 1.

Two types of C. citriodora trees (Large and Small) selected for this study

Methods

Plant materials

All plant materials were obtained from the Research Institute of Fast-growing Trees Chinese Academy of Forestry, and were identified by Professor Jianzhong Luo of the same institute, and are preserved at the South China Experiment Nursery (Suixi, Guangdong, China) with the deposition number: C. citriodora201509. Cambium of 6.5-year-old C. citriodora with large (B, 20.7–23.6 cm), medium (M, 14.8–15.8 cm), and small (S, 9.6–10.8 cm) diameters was collected at breast height (DBH) between 10 a.m. and 1 p.m. in winter. The DBH of different sizes were treated as three groups with six samples per treatment. All samples were collected and frozen in liquid nitrogen, then stored at − 80 °C until used.

Nontargeted metabolic profiling

The samples, consisting of six biological replicates, were dispatched to the Beijing Genomics Institute (Shenzhen, China). Comprehensive nontargeted metabolomic analysis was performed using a liquid chromatography-triple quadrupole mass spectrometry (LC-QQQ-MS) system in Multiple Reaction Monitoring (MRM) mode. The MRM transitions were established based on a self-constructed standard metabolite database, aligning the method with a ‘widely targeted’ metabolomics approach that enhances sensitivity and quantification for a predefined broad spectrum of metabolites. Metabolite identification was based on matching precise precursor/product ion pairs and retention times to authenticated standards in our database. Solid-phase extraction was performed as follows: Each sample was accurately weighed to 150 mg and transferred into a 2-mL thick-walled grinding tube containing 1 mL of extraction buffer and two small steel balls. The samples were subsequently homogenized using a tissue lyser at 50 Hz for 5 min. The homogenates were stored and vortexed every 10 min for three cycles, followed by an overnight incubation at 4 °C. Then, the samples were centrifuged at 13,000 g for 10 min at 4 °C. After centrifugation, 800 µL of the supernatant was filtered through a 0.22-µm filter membrane before liquid chromatography-mass spectrometry (LC-MS) analysis.

For liquid-phase extraction, 200 µL of each sample was mixed with 600 µL of extraction solvent. The mixture was then stored at -20 °C for a duration of 4 h or overnight. Following this, the samples were centrifuged at 20,000 g at 4 °C for 15 min, and the supernatant was collected. The supernatant was subsequently filtered through a 0.22-µm filter.

LC-MS analysis was conducted using a Waters UPLC I-Class Plus system (Waters, USA) coupled with a QTRAP 6500 Plus mass spectrometer (SCIEX, USA). Chromatographic separation was achieved with the following gradient elution program: 0–2 min, 5% mobile phase B (Acetone with 0.1% FA); 2–22 min, a linear gradient from 5% to 95% mobile phase B; 22–27 min, 95% mobile phase B; 27.1–30 min, re-equilibration to 5% mobile phase B. The flow rate was maintained at 0.3 mL/min, and the column temperature was set to 40 °C. For mass spectrometry, the QTRAP 6500 Plus was equipped with an EST Turbo Ion-Spray interface, and the source parameters were optimized as follows: source temperature, 450 °C; ion spray voltage (IS), 4500 V (positive mode) or -4500 V (negative mode); ion source gases I (GS1), II (GS2), and curtain gas (CUR) were adjusted to 40, 40, and 20 psi, respectively. Multiple reaction monitoring (MRM) methods were established in scheduled mode, incorporating specific MRM transitions, collision energy (CE), declustering potential (DP), and retention times for the targeted metabolites.

Upon acquisition of metabolome abundance data, subsequent analyses were conducted, including Principal Component Analysis (PCA) utilizing R software. Differentially accumulated metabolites (DAMs) were identified through a screening process with thresholds of |log2 fold change (FC)| ≥ 1 (fold change ≥ 2 or ≤ 0.5) and a significance level of p-value < 0.05 (without FDR adjustment due to the targeted nature of the metabolite assay), as previously described [25]. Pathway analysis for the identified DAMs was performed using the MetaboAnalyst online platform [26]. Additionally, the Mfuzz package was employed to cluster metabolites exhibiting similar expression patterns, thereby elucidating common trends in metabolic responses [27]. The Mfuzz package implements soft clustering to group genes/metabolites with similar expression/abundance trends across ordered samples, allowing for the identification of gradual increase or decrease patterns associated with tree size. The number of clusters was determined by evaluating cluster stability across a range of values, choosing the setting that provided biologically interpretable and distinct expression trends.

RNA extraction and illumina sequencing

Six independent biological RNA replicates obtained from three tree size (large, medium and small) were sequenced using a BGISEQ-500 system. Agilent 2100 Bioanalyzer (Agilent RNA 6000 Nano Kit) was employed for total RNA QC analysis. RNA purification, library construction, and RNA-seq were performed by the Beijing Genomics Institute (Shenzhen, China). The obtained reads were filtered using SOAPnuke software [28], and then the clean reads were compared to the reference genome sequence using HISAT [29]. The reads of each sequence tag were mapped to the reference using Bowtie2 [30]. After that, RSEM was used to calculate the expression levels of genes and transcripts [31], and DEGseq package was used for differential analysis [32]. DEGs were identified using the DEGseq package with a threshold of |log2FC| ≥ 1 and adjusted p-value (FDR) < 0.05. Eggnog, PlantTFDB, KEGG and GO databases were used to annotate genes. Based on Mfuzz package, the genes with the same expression trend were mined [27]. RNA-seq data newly generated in this study were deposited in GSA database (accession number: CRA035168).

Integrative analysis of growth traits and transcriptome

Phenotypic data for DBH and tree height (Table S1) were used for the initial WGCNA construction to explore broad gene-trait associations. However, given the primary focus of this study on radial growth, subsequent interpretation and candidate gene identification were centered on modules and genes significantly correlated with DBH. A gene co-expression network was constructed to elucidate the potential inter relationships between genes and growth attributes (DBH and Ht) via the Weighted Gene Co-expression Network Analysis (WGCNA) method, utilizing the R package (version 1.70) [33]. A signed network was constructed using a soft-thresholding power of 20 (selected based on scale-free topology fit), a minimum module size of 30, and a merge cut height of 0.25. Module-trait associations were calculated using Pearson correlation. Modules with significant correlations to the traits of interest were identified as candidate modules. To refine the list of candidate genes, the Gene Significance (GS) and Module Membership (MM) values were computed [34]. Hub genes within these candidate modules were then selected based on the criterion of an absolute MM value exceeding the set threshold (≥ 0.9). Hub genes within these candidate modules were selected based on high connectivity within the module (Module Membership, MM > 0.9) and strong correlation with the trait of interest (Gene Significance, GS > 0.2). These stringent thresholds are commonly used to identify high-confidence central regulators within co-expression networks. Subsequent to their identification, the protein sequences of the hub genes were subjected to KEGG and Gene Ontology (GO) enrichment analyses for functional insights. Pathways with a p-value (or FDR-adjusted q-value) < 0.05 were considered significantly enriched.

Results

Overview of molecular differences between tree size classes

Principal component analysis (PCA) found significant separation between S and B and M (Fig. 2A), indicating that metabolome test results can be used for differential metabolite screening analysis. Based on the self-established database, qualitative and quantitative analysis of the metabolites in the samples was carried out. A total of 518 metabolites were detected through the triple quadrupole (QQQ) mass spectrometer screening (Table S2). These metabolites can be divided into 30 classes, of which flavonoid metabolites are the most numerous (Fig. 2B).

Fig. 2.

Metabolome analysis of three different stem girths of Corymbia citriodora. A Principal component analysis of trees of different sizes. B Classification and proportion of metabolites. C-E Volcanic maps of differential metabolites in different comparison groups. S: Small DBH; M: Medium DBH; B: Large DBH. Red dots: upregulated DAMs; blue dots: downregulated DAMs; gray dots: non-DAMs

Based on the screening criteria of |log2FC| ≥ 1 and p-value < 0.05, a total of 51, 102, and 68 DAMs were identified in the comparisons of B vs. M, S vs. B, and S vs. M, respectively. In the B vs. M comparison, 17 metabolites were downregulated in B, while 34 were upregulated. Among the upregulated metabolites, the one with the highest fold change was 1,2,3,4,6-pentagalloylglucose, followed by typhaneoside and picroside I (Fig. 2C). In the S vs. B comparison, 49 metabolites were downregulated in S, and 53 were upregulated. The metabolite with the greatest fold change among the downregulated compounds was also 1,2,3,4,6-pentagalloylglucose (Fig. 2D). In the S vs. M comparison, 21 metabolites were downregulated in S, and 47 were upregulated. The metabolite with the highest fold change among the downregulated compounds was picroside I, followed by orientin and delsoline (Fig. 2E). These results indicate that phenolic compounds, particularly tannins like 1,2,3,4,6-pentagalloylglucose, are major differential metabolites in the cambium of B trees.

A total of 14 metabolites up-regulated in B and M relative to S were found, among which 4 metabolites (picroside I, orientin, delsoline, corilagin) were up-regulated in B relative to M (that is, the intersection of metabolites in S, M and B) (Figure S1A).

KEGG enrichment analysis was conducted on DAMs. The results revealed that in the B vs. M comparison, there were 5 pathways (map00943: Isoquinoline alkaloid biosynthesis, 00944: Flavonoid biosynthesis, 00999: Various secondary metabolites, 01232: Nucleotide metabolism, 00010: Glycolysis) with a P-value less than 0.05, and 12 metabolites were annotated to these 5 pathways (Figure S1B). Similarly, in the S vs. B comparison, there were also 5 pathways (map00943, 00944, 00999, 00470: D-Amino acid metabolism, 00052: Galactose metabolism) with a P-value less than 0.05, with 14 metabolites annotated to these pathways (Figure S1C). In the S vs. M comparison, there were five pathways (map00941: Stilbenoid biosynthesis, 00330: Arginine and proline metabolism, 02010: ABC transporters, 00230: Purine metabolism, 01232) with a P-value less than 0.05, and 9 metabolites were annotated to these pathways (Figure S1D).

In addition, to identify metabolites whose abundance correlates with increasing DBH, the metabolites in the samples of different groups were clustered using Mfuzz, and 38 metabolites were identified in Class 6 that showed a progressive increase in abundance from S to M to B trees (Figure S2), suggesting a potential association with radial growth.

Divergent phytohormone signaling suggests differential stress responses

To delve into the molecular underpinnings of wood formation, an RNA-seq analysis was conducted on C. citriodora with varying DBH. A total of 18 samples were sequenced, each generating an average of 21.59 million reads. Following the acquisition of clean reads, these were aligned to the reference genome sequence using HISAT. The results revealed an average alignment rate of 90.38% to the genome and 85.58% to the gene set across the samples. A total of 30,970 genes were detected, which were subsequently subjected to functional annotation. Of these, 13,694 (44.22%) genes were annotated in the GO database, while 13,964 (45.09%) were annotated in the KEGG database, and 1,925 genes were identified in the PlantTFDB (Table S3). Based on the criteria of |log2 fold change (FC)| ≥ 1 and FDR-adjusted p-values ≤ 0.05, 5,336 DEGs were identified in the S versus B comparison, of which 2,587 (48.48%) were upregulated in the B group (Table S4). In the S vs. M comparison, 4,065 DEGs were identified, with 1,664 (40.93%) upregulated in the M group (Table S5). Additionally, 2,583 DEGs were found in the M vs. B comparison, with 1,391 (53.85%) upregulated in the B group (Table S6). To investigate transcriptional trends associated with increasing DBH, gene expression levels were clustered across different sample groups. Clusters 3, 12, and 8 exhibited an upward trend from S to B trees and may be associated with the increase in tree stem girth (Figure S3).

GO annotation and KEGG pathway analysis were used to identify the biological functions of DEGs. GO functional enrichment analysis was performed for DEGs in each comparison group (Fig. 3A-C). The results revealed that most DEGs significantly enriched in ‘component of membrane’ in three comparison group, while the DEGs related to ‘cell cycle’ was significantly enriched in S vs. B. The KEGG enrichment of DEGs reflected that the differential genes in the three comparison groups were mainly related to starch and sucrose metabolism, plant-pathogen interactions, galactose metabolism, protein processing in the endoplasmic reticulum, and ABC transporters (Fig. 3D-F).

Fig. 3.

GO annotation and KEGG pathway analysis. A-C The GO function annotation of differential genes in B vs. M, S vs. B and S vs. M. D-F The KEGG function annotation of differential genes in B vs. M, S vs. B and S vs. M. S: Small DBH; M: Medium DBH; B: Large DBH

To determine whether differential stress sensitivity underlies the observed growth variation, we examined phytohormone-related transcripts. Within the phytohormone signaling pathways, 22 differentially expressed genes were identified. Genes related to ABA synthesis (NCED) and signaling (PYR/PYL receptors, ABF transcription factors) were predominantly upregulated in small trees (Fig. 4A). Specifically, Cocit.A1269, Cocit.D1392, Cocit.E0025 (PYR/PYL) and Cocit.L4711 (ABF) showed significant upregulation in S trees. Additionally, MYC2, which interacts with ABF to modulate stress responses, was upregulated in small trees. In contrast, six genes associated with auxin induction were upregulated in large trees, while most gibberellin-related genes showed downregulation in B trees (Fig. 4A). This coordinated upregulation of ABA pathway components in S trees suggests an active stress response that may inhibit growth, whereas B trees maintain a transcriptional profile permissive for continued development.

Fig. 4.

Transcriptomic profiling of various metabolism pathway related to wood formation. A Genes involved in plant hormone signal transduction. B Genes involved in starch and sucrose metabolism. C Pentose and glucuronate interconversions. D Cell cycle-related genes. E Phenylpropanoid biosynthesis. F Phenylalanine, tyrosine and tryptophan biosynthesis. Red: high expression; blue: low expression. For detailed color legend, see within the figure

Lignin biosynthesis pathway reveals contrasting regulatory strategies

To investigate whether cell wall regulation differs between tree size classes, we examined genes and metabolites involved in lignin biosynthesis. Lignin is a pivotal component of the plant cell wall. In the lignin biosynthetic pathway (Fig. 5), we identified 14 metabolites, of which shikimic acid, 4-hydroxyphenylacetic acid, gentisic acid, cinnamic acid and syringin showed increased abundance with increasing stem girth, whereas the abundance of tyrosine decreased. Integrating transcriptomic data, we found that shikimic acid is regulated by shikimate dehydrogenase (SDH, [EC:1.1.1.25]), primarily controlled by the gene Cocit.H2978, which was significantly upregulated in the large tree. Cinnamic acid is under the regulation of phenylalanine ammonia-lyase (PAL, [EC:4.3.1.24]) and phenylalanine/tyrosine ammonia-lyase (PTAL, [EC:4.3.1.25]), influenced by Cocit.A1965 and Cocit.D1995, respectively. Cocit.A1965 was downregulated, while Cocit.D1995 was upregulated in the large tree. DEGs involved in tyrosine metabolism (Cocit.A0510/ASP5 and Cocit.K1525) were upregulated in the large tree.

Fig. 5.

Metabolic and genetic pathways associated with lignin synthesis. The red font represents the detected metabolites. SDH: shikimate dehydrogenase; APS5: aspartate aminotransferase; TYDC: tyrosine decarboxylase; AOC: primary-amine oxidase; PAL: phenylalanine ammonia-lyase; PTAL: phenylalanine/tyrosine ammonia-lyase; 4CL: 4-coumarate–CoA ligase; HCT: shikimate O-hydroxycinnamoyltransferase; CYP98A: 5-O-(4-coumaroyl)-D-quinate 3’-monooxygenase; CCOAMT: caffeoyl-CoA O-methyltransferase; CCR: cinnamoyl-CoA reductase; CAD: cinnamyl-alcohol dehydrogenase; PRO: peroxidase

While shikimic acid and SDH were both elevated in B trees—consistent with sustained precursor supply—the downstream CAD genes were uniformly upregulated in S trees (9/9, 100%). Furthermore, seven genes (7/12, 58%) encoding peroxidases (PRO) were significantly upregulated in the small tree. This inverse relationship suggests that S trees may channel available precursors toward polymerization, whereas B trees maintain a metabolic state permissive for continued cell expansion with delayed lignification.

To identify genes potentially involved in cell wall dynamics, we screened genes harboring domains associated with cellulose synthesis (Cellulose_synt). From 33 such genes, 10 were differentially expressed in the S vs. B comparison. Focusing on genes highly expressed in B trees (FPKM > 5 in B group), we selected 14 genes as candidate regulators. Among these, the majority (9 out of 11) of genes related to primary cell wall (PCW) biosynthesis exhibited higher expression levels in the B group, suggesting more active cell division and expansion. In contrast, genes related to secondary cell wall (SCW) synthesis (Cocit.C0102, Cocit.A0286, Cocit.D0255) had relatively higher expression levels in the S group, indicating a transcriptional shift towards SCW formation and potential lignification in smaller trees (Figure S4). Furthermore, from the genes upregulated in B trees, we identified 25 potential regulators containing domains such as PAE, PMEI, Glyco_hydro_16, XET_C, Pectinesterase, Glyco_transf_92, which are involved in pectin modification and hemicellulose metabolism. Additionally, among downregulated genes, eight containing domains including Tryp_alpha_amyl, Glyco_hydro_16, XET_C, PMEI, Pectinesterase, and Glyco_transf_43 may have contrasting roles.

To pinpoint genes uniquely expressed in large trees, we identified B-specific genes in cambium (FPKM < 1 in S and M, > 5 in B). A total of 11 genes were identified, among which 5 were endochitinase and 1 was adenosylhomocysteinase. Two belong to the cytochrome P450 family.

Carbohydrate metabolism reflects distinct carbon allocation strategies

To explore whether carbon partitioning differs between tree size classes, we analyzed starch and sucrose metabolism pathways. In the starch and sucrose metabolism pathway, the majority of genes were upregulated (18/20) in B trees. Notably, sucrose synthase (SUS) was upregulated (Fig. 4B), indicating potential for abundant monosaccharide provision for biosynthesis. Alpha-amylase (AMY), which breaks down starch, was also upregulated in the B tree, implying continued mobilization of carbohydrate reserves.

In pentose and glucuronate interconversions, 11 out of 13 genes were upregulated in the B tree, predominantly pectate lyase (PEL), which may facilitate cell wall remodeling. Polygalacturonase (PGA) and UDP-glucose 6-dehydrogenase were also upregulated (Fig. 4C).

The abundance of raffinose was found to be higher in the small tree. However, most genes associated with raffinose synthesis (RFS) showed no significant differences between tree classes, suggesting post-transcriptional regulation or differential turnover. Manninotriose was also found to be higher in the small tree. As an important carbohydrate, manninotriose may not only serve as a storage form of energy and carbon sources but also play a significant role in protecting cell structure, resisting oxidative stress, and regulating cell osmotic pressure.

Furthermore, genes related to trehalose metabolism (TPP) showed a distinct pattern, with some (Cocit.I0363 and Cocit.I0364) significantly upregulated in B trees. A total of nine TPP genes were identified, but only these two showed significant upregulation in large trees. Taken together with the upregulation of SUS and AMY, B trees appear to maintain a metabolism geared toward sucrose production and utilization, potentially supporting continued biosynthetic processes, while S trees accumulate specific protective oligosaccharides like raffinose.

Given the strong enrichment for galactose metabolism, we constructed a network linking galactose-related metabolites with DEGs (Fig. 6). The results indicate that the synthesis or metabolism of raffinose is influenced by raffinose synthase ([EC:2.4.1.82]) and alpha-galactosidase ([EC:3.2.1.22]), with significant upregulation of Cocit.F3043 and Cocit.G2043 in the B tree. The synthesis and metabolism of sucrose are regulated by maltase ([EC:3.2.1.20]), sucrose synthase ([EC:2.4.1.13]), and alpha-galactosidase, with significant upregulation of Cocit.L3962 (malZ), Cocit.G2043 (GLA), Cocit.F2680 (SUS), Cocit.F2676 (SUS), and Cocit.J1034 (SUS) in the B tree. Moreover, enzymes involved in cellulose degradation, endoglucanase ([EC:3.2.1.4]) and beta-glucosidase ([EC:3.2.1.21]), were generally upregulated in the B tree, suggesting active cell wall turnover.

Fig. 6.

Network of differential gene and galactose-related metabolite pathways. The red font indicates metabolites detected by the metabolome. The green letters represent enzymes and their corresponding gene. GLAM: aldose 1-epimerase; GOLS: inositol 3-alpha-galactosyltransferase; RFS: raffinose synthase; EG: endoglucanase; BG: beta-glucosidase; TPP: trehalose 6-phosoatase; GLA: alpha-galactosidase; INV: beta-fructofuranosidase; SUS: sucrose synthase; TPS: trehalose 6-phosphate synthase

Cell-cycle and cytoskeletal genes support sustained cambial activity in large trees

To determine whether cambial cell division differs between tree size classes, we examined cell-cycle regulatory genes. Consistent with potential sustained cambial activity, most genes regulating the cell cycle (12/15) were upregulated in the B tree, with Mini-Chromosome Maintenance (MCM) genes playing a dominant role (Fig. 4D). Nine MCM gene family members were identified in C. citriodora, of which six MCM genes were upregulated in the large tree, including those homologous to Arabidopsis MCM4, MCM5, and MCM6. Their expression supports the notion of sustained cambial cell-cycle activity in B trees during winter.

In phenylpropanoid biosynthesis, enzymes like 4-coumarate-CoA ligase (4CL) and cinnamoyl-CoA reductase (CCR) were upregulated, while three cinnamyl-alcohol dehydrogenase (CAD) genes were downregulated in the B tree (Fig. 4E), pointing to differential regulation of lignin monomer synthesis. In the phenylalanine and tyrosine metabolic pathway, shikimate dehydrogenase (aroE/SDH) and aspartate aminotransferase (ASP5) were upregulated in the B tree (Fig. 4F), suggesting enhanced flux into the aromatic amino acid pathway.

Co-expression network identifies hub genes associated with DBH

To uncover genes associated with growth traits, we conducted WGCNA based on gene expression levels and growth data. Phenotypic data for DBH and tree height were used for the initial construction to explore broad gene-trait associations. Given the primary focus on radial growth, subsequent interpretation centered on DBH. The results indicated a significant positive correlation between the genes within the ‘antiquewhite4’ module and growth traits (Figure S5A). This module was enriched for processes like ‘DNA replication’ and ‘microtubule-based movement’. To refine the gene set within the antiquewhite4 module, we performed MM-GS analysis on its 4,584 genes, identifying 499 genes with MM > 0.9 and GS > 0.2 (Figure S5B). Among these, 27 transcription factors were identified, the majority of which were upregulated in the B tree. Notably, the ethylene response factor (ERF) was upregulated in the S tree (Figure S5C). The gene with the highest MM value was Cocit.H0106 (a kinesin, KIN12B). GO and KEGG analyses revealed that genes promoting wood formation were predominantly enriched in pathways related to DNA replication, cell division, and microtubule activity (Figure S5D-E).

The antiquewhite4 module, significantly correlated with DBH, provides independent support for sustained cambial activity in B trees. This module’s enrichment in DNA replication and microtubule-based movement genes aligns with the DEG-based observation of enhanced cell-cycle gene expression. Critically, the module hub gene Cocit.H0106 (KIN12B)—a kinesin involved in microtubule organization—connects cell division machinery with cytoskeletal dynamics required for expanding cambial cells. The inclusion of MYB and E2F3 among high-MM genes suggests these transcription factors may orchestrate the module’s activity. Thus, WGCNA not only reinforces the DEG findings but also identifies potential regulatory hierarchies: E2F3 and MYB as upstream regulators, MCM and kinesin as downstream effectors of sustained cambial activity.

Discussion

Wood development is influenced by numerous factors. C. citriodora wood, which serves a variety of purposes, naturally garners higher economic benefits with increased yield. To uncover the potential mechanisms underlying differences in wood size under winter conditions, we conducted integrated omics analyses. Our data support the hypothesis of divergent adaptation strategies: small (S) trees exhibit molecular signatures of a heightened stress response and early preparation for dormancy, while large (B) trees maintain transcriptional and metabolic profiles conducive to sustained cambial activity and growth.

Differential dormancy onset as a primary driver of DBH variation

A central finding of this study is that S and B trees exhibit fundamentally different phytohormone signatures during winter, suggesting that differential timing of dormancy onset may explain the observed variation in DBH. Genes related to ABA synthesis (NCED) and signaling (PYR/PYL, ABF, MYC2) were upregulated in S trees. ABA plays a pivotal role in wood development, particularly in modulating secondary cell wall formation and lignin deposition [35], and facilitates dormancy while inhibiting growth under stress conditions [36, 37]. The coordinated upregulation of these genes in S trees suggests an active ABA-mediated stress response pathway that could inhibit growth and accelerate dormancy onset. While our interpretations are based on transcript levels, direct measurements of endogenous phytohormones would strengthen these conclusions.

Carbon partitioning reflects alternative survival strategies

Contrasting patterns of carbohydrate metabolism reveal distinct carbon allocation strategies between the two size classes. S trees accumulated raffinose and manninotriose, which are known for their roles in stress protection and osmoregulation [38, 39], without corresponding upregulation of their biosynthesis genes. This suggests post-transcriptional regulation or reduced turnover, indicating a shift toward stress protection. Conversely, B trees upregulated genes involved in sucrose synthesis and mobilization (SUS, TPP, AMY), reflecting a metabolism geared toward continued biosynthesis. Trehalose, regulated by TPP genes, acts as an osmoprotectant in higher plants [40], and the selective upregulation of specific TPP genes in B trees suggests a fine-tuned protective metabolism that does not preclude growth. Thus, while S trees prioritize stress protection, B trees maintain a growth-compatible metabolic state.

Cell-wall regulation distinguishes permissive vs. restrictive growth states

The contrasting patterns of primary and secondary cell wall gene expression between the two size classes suggest that the timing of lignification gates the window for radial expansion. B trees were characterized by an enrichment of primary cell wall genes (CESA), cell-cycle regulators (MCM, E2F3), and the lignin precursor shikimic acid, along with its biosynthetic gene SDH [41]. This indicates a permissive state for cell division and expansion. In contrast, S trees showed an upregulation of genes involved in later-stage lignin polymerization (CAD, PRO peroxidases), suggesting an accelerated lignification program that likely closes the window for radial expansion under cold stress. Additionally, the higher abundance of the ellagitannins PGG and corilagin in the cambium of B trees suggests a potential link between specialized metabolism and sustained radial growth, consistent with their known roles as antioxidants and defensive compounds in Eucalyptus [42, 43].

Coordinated regulation of cambial activity by hub genes

WGCNA identified a core gene set associated with DBH, providing insight into regulatory hierarchies controlling cambial activity. In B trees, this set included E2F3, a key regulator of cell cycle progression, and six MCM genes. In poplar, E2F transcription factors regulate cell-cycle genes including MCMs and cyclins, and are crucial for wood development [44]. Under low-temperature stress, MCM homologs are upregulated in Brassica species [45], supporting a model wherein sustained cambial cell-cycle activity occurs even under winter conditions in B trees. The upregulation of multiple kinesin genes in B trees further aligns with active cytoplasmic organization during growth [46]. Conversely, genes associated with secondary cell wall synthesis, including CESA and a C2C2-dof transcription factor [47], were downregulated in B trees, likely facilitating cell proliferation by delaying the lignification program.

Integrated model: divergent winter adaptation strategies mediate wood formation

Synthesizing these findings, we propose a unified framework where differential dormancy timing explains within-family growth variation (Fig. 7). Under freezing stress, S trees activate an ABA-driven stress response and accumulate raffinose-family oligosaccharides, leading to accelerated lignification and a restrictive state for growth. In contrast, B trees upregulate genes for sucrose metabolism and cell division (E2F3, MCM), maintain a permissive cell wall state, and adjust cell wall physical properties, allowing for sustained cambial activity. These patterns align with previous observations that differentially expressed genes and metabolites between cold-hardy and cold-sensitive individuals are primarily enriched in starch, sucrose metabolism, and phenylalanine metabolism [48]. Our multi-omics data reveal three patterns of transcript-metabolite coordination: (i) concordant upregulation (e.g., SDH-shikimic acid), (ii) discordant regulation suggestive of post-transcriptional control (e.g., raffinose accumulation), and (iii) pathway-level shifts (e.g., phenylpropanoid pathway), indicating that DBH differences emerge from systems-level reorganization.

Fig. 7.

Hypothesized model of divergent winter adaptation strategies mediating wood formation in Corymbia citriodora trees of varying diameters. The model contrasts the inferred molecular and metabolic states in Small (S) and Large (B) trees under winter conditions. Solid lines/boxes indicate relationships or processes strongly supported by differential expression/accumulation data in this study. Dashed lines/boxes represent proposed interactions based on established literature

Collectively, our multi-omics approach provides compelling correlative evidence for divergent winter adaptation strategies in C. citriodora. The functional roles of identified candidate genes and the causal relationships within our proposed model require future validation through genetic, physiological, and anatomical studies. For instance, transgenic approaches targeting E2F3 or CAD in cambium could directly test their roles in regulating the timing of growth cessation and radial expansion.

Conclusions

This study advances understanding of tree growth variation by demonstrating that winter conditions do not uniformly suppress cambial activity but instead reveal divergent adaptation strategies linked to DBH. The key conceptual advance is the integration of stress signaling, carbon metabolism, and cell-wall regulation into a unified framework where differential dormancy timing explains within-family growth variation. Unlike model species studies that identify individual regulators, our multi-omics approach reveals that large trees maintain a permissive growth state through coordinated upregulation of cell-cycle, cytoskeletal, and biosynthetic genes. Conversely, small trees exhibit coordinated stress responses that may accelerate dormancy. These findings provide species-specific targets for functional validation and suggest that breeding for sustained winter cambial activity could enhance biomass production in subtropical plantations. More broadly, this work establishes that within-family DBH variation in forest trees can be explained by differential environmental responses rather than solely genetic determinism, with implications for both breeding strategies and our understanding of tree adaptation to seasonal environments.

Authors’ contributions

WC wrote the manuscript and analyzed the data. LX and SZ conducted experiments and acquired the data. HW and OY conceived the research framework, spearheaded experimental design, supervised methodological implementation, and critically edited the manuscript. All authors read and approved the final version of the manuscript.

Funding

Supported by program for scientific research start-upfunds of Guangdong Ocean University– Breeding and variety expansion of wind resistant and salt tolerant Eucalyptus camaldulensis in coastal windbreak zones (No.060302052403); Zhanjiang Science and Technology Plan Program (No. 2023A01026).

Data availability

RNA-seq data newly generated in this study were deposited in GSA database (accession number: CRA035168).

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not Applicable.

Competing interests

The authors declare no competing interests.