The experimental material was ‘Chuan Hou Po’ bark from Magnolia officinalis Rehd. et Wils., grown in Sichuan Province. Samples were collected from dried stem bark from ten-year-old (DR) and two-year-old (XR) Magnolia officinalis trees in Wahugou Village, Beichuan Qiang Autonomous County, Mianyang City, Sichuan Province (Coordinates: 31°58^′46.38^″ N, 104°30^′39.00^″ E). Trees were categorized depending on age. Three trees per age group (DR: DR1, DR2, DR3; XR: XR1, XR2, XR3) were sampled to create biological duplicates.

The stem epidermis of Magnolia officinalis Cortex was frozen in liquid nitrogen and delivered to Major Biotechnology Co., Ltd. for second-generation transcriptome sequencing. After removing junctions and low-quality readings, clean data was generated. According to the scaffold-level Magnolia officinalis Cortex genome sequence provided by Chengdu University of Traditional Chinese Medicine, the full coding gene sequence of Magnolia officinalis Cortex was obtained via sequence splicing and annotated in the database. The NCBI gene database (https://www.ncbi.nlm.nih.gov/gene/, accessed on 07 June 2025) was utilized to collect protein sequence data for Magnolia officinalis. 135 Arabidopsis AP2/ERF protein sequences were obtained from the TAIR database (https://www.arabidopsis.org/index.jsp, accessed on 07 June 2025).

The protein sequences of Magnolia officinalis were examined and evaluated using the local BLAST software against those of 135 Arabidopsis AP2/ERF gene family members. To identify high-reliability candidate sequences, we used the following criteria: E-value < 1 × 10⁻¹⁰, comparison score > 100, and a single terminator. Second, the information acquired in the previous stage is summarized, duplicate data is deleted, and redundant sequences are eliminated. The possible AP2/ERF sequence is next examined to check if it contains the totally preserved AP2 domain using the SMART online software (http://smart.embl-heidelberg.de/, accessed on 07 June 2025) and the InterProScan online software. The discovered protein sequence, which comprised one or two conserved AP2 domains, was retained and identified as the AP2/ERF gene family of Magnolia officinalis. Furthermore, these AP2/ERF gene families were categorized based on the conserved AP2 domains of the AP2/ERF gene subfamily.

Protparam (https://www.expasy.org/resources/protparam, accessed on 07 June 2025) was used in the online analysis software Expasy to determine the length of the encoded amino acid, the total number of positive and negative charge residues, molecular weight (MW), theoretical isoelectric point (pI), wobble factor, aliphatic index, and grand average of hydropathicity (GRAVY). The subcellular localization of the Magnolia officinalis AP2/ERF protein sequence was then carried out utilizing the web platform Cell-PLoc 2.0 (Plant-mPLoc, http://www.csbio.sjtu.edu.cn/bioinf/Cell-PLoc-2/, accessed on 07 June 2025).

The 135 Arabidopsis AP2/ERF protein sequences, as well as the previously reported Magnolia officinalis AP2/ERF gene family protein sequences, were loaded into MEGA11.0 software to create a phylogenetic tree. After importing all of the sequences, Align by ClustalW was used for multiple sequence alignment, and the Neighbor-Joining method was used to build the phylogenetic tree of the Magnolia officinalis AP2/ERF and Arabidopsis AP2/ERF gene families. The AP2/ERF protein motif was predicted using the internet software MEME (http://meme-suite.org/tools/meme, accessed on 07 June 2025). The number of motif predictions was fixed to ten, with a minimum length of six amino acids and a maximum of fifty.

The Prabi (https://npsa-prabi.ibcp.fr/cgi-bin/npsa_automat.pl?page=npsa_sopma.html, accessed on 07 June 2025) was used to access the secondary structure prediction module, and SOPMA was selected to predict the secondary structure of the Magnolia officinalis AP2/ERF protein. The AP2/ERF protein sequence of Magnolia officinalis was analyzed using the SWISS-MODEL online software (https://www.swissmodel.expasy.org, accessed on 07 June 2025). The model was created using the database template with the greatest resemblance, and the tertiary structure of the Magnolia officinalis AP2/ERF protein was determined.

Magnolia officinalis RNA-Seq data (relevant data can be found in Supplementary Materials) from our previous study were divided into two groups: Magnolia officinalis at mature growth (DR1, DR2, DR3) and Magnolia officinalis at early growth (XR1, XR2, XR3). The gene expression of the AP2/ERF gene family in Magnolia officinalis was examined during the two groups’ respective growth periods, and a heatmap tree was created. The AP2/ERF genes that were overexpressed in the two groups were annotated in GO (https://www.geneontology.org/) and KEGG (https://www.genome.jp/kegg/, accessed on 07 June 2025).

The gene-level expression data were used in shapiro.test() and levene Test() to perform a normal distribution test and a homogeneity of variance test in R. When the test results showed normal distribution (p-value > 0.05) and homogeneity of variance (p-value > 0.05), t.test() was used; otherwise, wilcox.test() was used to obtain two groups of Magnolia officinalis AP2/ERF genes with significant difference (p-value < 0.05) in expression during different growth periods. These differential genes were phylogenetically matched with AP2/ERF homologous genes similar to Arabidopsis thaliana, and structural and functional prediction analysis was performed using the AlphaFold Protein Structure Database (https://alphafold.com).

The protein sequence of Magnolia officinalis was compared to the protein sequence of 135 members of the AP2/ERF gene family in Arabidopsis thaliana, yielding 11553 protein sequences. After the information summary, we deleted repetitive items, sequences with E-values larger than 1 × 10⁻¹⁰, and those with comparison scores less than 100. Following that, 89 protein sequences were retrieved, and proteins with two or more terminators and the same protein sequence were excluded. The protein sequence containing the AP2 domain was tested for redundancy using the SMART network online tools. 75 AP2/ERF sequences with the AP2 domain were identified and called MoAP2/ERF1–MoAP2/ERF75. The 75 MoAP2/ERF1–MoAP2/ERF75 were divided into 4 subfamilies (Table 1): 18 AP2 members (2 AP2 domains) [33], 5 Soloist members (AP2 conserved domains have more differences) [13], 18 DREB members (1 AP2 domain; the 14th and 19th amino acids are valine and glutamic acid, respectively) and 34 ERF members (1 AP2 domain; the 14th and 19th amino acids are alanine and aspartic acid, respectively) [34,35]. There was no B3 domain found in the AP2/ERF gene family of Magnolia officinalis reported in this study, implying that no AP2/ERF gene belonged to the RAV subfamily [36].

We used Expasy to determine the physicochemical parameters of these 75 MoAP2/ERF protein sequences (Table 1). The results revealed that the amino acid chain length of most MoAP2/ERF proteins ranged from 72 to 664 aa, with an average of 306 aa and an average relative molecular weight of around 33,779.84 Da. MoAP2/ERF51 has the smallest amino acid length of 72 aa, whereas MoAP2/ERF57 has the greatest amino acid length of 664 aa, despite being members of the same AP2 subfamily. This demonstrates that the length and size of AP2/ERF gene family proteins in Magnolia officinalis vary greatly. The theoretical isoelectric point of the MoAP2/ERF protein is 4.83–9.65, indicating a large degree of variability in the acid-base range. The proportion of theoretical isoelectric points less than 7.00 is relatively high, about 62% (29), indicating that the number of acidic MoAP2/ERF proteins is greater than that of alkaline MoAP2/ERF proteins, and most of the DREB subfamily proteins are acidic (except MoAP2/ERF31, MoAP2/ERF40, and MoAP2/ERF73). In terms of the protein stability study, practically all proteins had an instability coefficient of more than 40, indicating that they were unstable, with only 9 MoAP2/ERF proteins remaining stable. The protein aliphatic index ranged between 46.31 and 76.84. The total average hydrophilicity of the MoAP2/ERF proteins was negative, and they were all hydrophilic. Using subcellular localization results (Table 1), MoAP2/ERF proteins were primarily localized in the nucleus (56), with a minor number of proteins localized in the cytoplasm (6) and others able to shuttle between the nucleus and the cytoplasm (13). It can be seen that the MoAP2/ERF gene family is primarily found in the nucleus and performs an active function.

To generate a neighbor-joining phylogenetic tree, the MoAP2/ERF protein sequence was compared to the AP2/ERF protein sequence of the model plant Arabidopsis thaliana using MEGA11.0 software (Fig. 1). The results indicated that the phylogenetic tree was divided into four subfamilies: AP2, Soloist, ERF, and DREB. MoAP2/ERF in the ERF subfamily and AP2/ERF in Arabidopsis thaliana have additional branches, showing that these ERF subfamilies’ activities are complex and changeable. Furthermore, we identified that one branch of the DREB subfamily (which includes MoAP2/ERF15, MoAP2/ERF25, MoAP2/ERF27, and MoAP2/ERF28) would be more comparable to the Soloist subfamily during differentiation. The online analytic software MEME was used to predict and analyze the conserved motifs in the MoAP2/ERF protein sequence, and the results are shown in the Figs. 2 and 3. Almost all MoAP2/ERF protein sequences contain Motif1, Motif2, Motif3, and Motif4. These are processed via α-helix, β-sheet, extension, and random coil to form the conserved AP2 domain. The AP2 domain is formed in the following order: Motif 2, Motif 4, Motif 1, and Motif 3, followed by Motifs 8 and 5. In addition to the AP2 domain, the conserved motifs show that the majority of MoAP2/ERF proteins are extremely conserved. Their delicate structure and high amount of differentiation make other amino acid sites less predictable.

Plants’ AP2/ERF domain consists of three anti-parallel β-sheets, a parallel α-helix, and random coils [37]. The MoAP2/ERF gene family’s secondary structure prediction (Table 2) and structural difference analysis (Fig. 4) identified numerous α-helix and random coil proteins. MoAP2/ERF30 contains a substantial amount of α-helix, with the rest ranging from 20.00% to 40.00%, indicating a highly conserved AP2 domain. The fraction of irregular curls spans from 32.35% to 71.07% and is the largest. Furthermore, the proportion of random coils in MoAP2/ERF30 and MoAP2/ERF51 is low, indicating that MoAP2/ERF30 has a more conservative and regular secondary structure. The proportion of β-sheets was the least, whereas long chains were concentrated in 2.73%–17.41%. The majority of MoAP2/ERF gene family proteins in the SWISS-MODEL online model have a Seq-identity of at least 70%. Seven MoAP2/ERF gene family proteins share the ERF structural model of Gcc-box binding. There are 32 ethylene-responsive transcription factors or ERF crystal structure models, 31 AP2/ERF-related protein models, and CBF-like transcription factor models (Fig. 5).

The heatmap tree of 75 MoAP2/ERF gene families was constructed in the order of DREB to ERF utilizing the gene expression levels of Magnolia officinalis throughout the early growth period (XR1, XR2, XR3) and the mature growth period (DR1, DR2, DR3) (Fig. 6). MoAP2/ERF31 and MoAP2/ERF32, which share a high degree of similarity, were overexpressed in all samples. The expression of MoAP2/ERF genes was higher in XR than in DR. MoAP2/ERF62 and MoAP2/ERF70, with substantial homology, as well as the last secondary branch, MoAP2/ERF66–MoAP2/ERF58, demonstrated that XR expression was higher than that of DR. Additionally, MoAP2/ERF71 and MoAP2/ERF69 of the Soloist subfamily have obvious expression ability on XR and DR. In the ERF subfamily, there are more active AP2/ERF genes, such as MoAP2/ERF39, MoAP2/ERF36, MoAP2/ERF37, MoAP2/ERF55, MoAP2/ERF45, MoAP2/ERF56, MoAP2/ERF60, MoAP2/ERF22, MoAP2/ERF20, MoAP2/ERF1, and MoAP2/ERF43, which are substantially expressed in both DR and XR.

We performed GO and KEGG functional annotations on these 15 highly expressed MoAP2/ERF genes and found functional classifications for 11 of them in the GO database, including biological process, cellular component, and molecular function (Fig. 7). Among these, MoAP2/ERF22 and MoAP2/ERF1 play biological regulatory roles. MoAP2/ERF22, MoAP2/ERF60, MoAP2/ERF32, MoAP2/ERF71, and MoAP2/ERF1 are all responsive to stimuli. MoAP2/ERF55 can activate imidazole glycerol phosphate synthase, which is involved in histidine biosynthesis. In vivo, MoAP2/ERF39 acts as a transcriptional activator for Pti6, causing the activation of multiple PR genes with the GCC-box and performing a critical and unique function in plant defense.

We examined the gene expression levels of the MoAP2/ERF gene family throughout two growth periods (XR and DR) and discovered 15 substantially changed genes (Fig. 8). Ten of these differentially expressed genes have three functions: DNA binding, transcription regulator activity, and gene expression control, with the remaining five also involved in stress response and primary metabolic processes. This suggests that they often influence Magnolia officinalis gene expression by binding to DNA, affecting the plant’s ability to tolerate adverse environmental variables and create secondary metabolites. In some areas, XR may outperform DR.