The seeds were collected in October from Minhou County, Fuzhou City, Fujian Province (26°4′39.48″N, 119°10′54.27″E). Following pulp removal, the freshly harvested seeds were rinsed and immediately underwent an after-ripening process through cold stratification at 4 °C in moist sand for approximately 120 days. This moist-chilling treatment was applied to break physiological dormancy and promote metabolic preparation for germination. After stratification, the seeds were transferred to a growth chamber set at 25 °C with a 14/10 h light/dark cycle to induce germination. Germination progression and seedling development were monitored every 5 days. To elucidate the germination dynamics, three critical stages were defined. At each stage, biological replicates (n=10 seeds per sample) were randomly harvested, quenched in liquid nitrogen and stored at −80 °C for subsequent experiment and analysis. Three independent biological replicates were performed in this study.

Total RNA was extracted from seeds using the Trizol method. The concentration and integrity of the total RNA were assessed. Total RNA was used for library preparation following the manufacturer’s protocol (New England Biolabs, Ipswich, MA, USA). All nine libraries, representing three different stages with triplicate biological replicates designated as S1-1, S1-2, S1-3; S2-1, S2-2, S2-3; S3-1, S3-2, S3-3, were sequenced on the Illumina HiSeq X Ten platform (Illumina, San Diego, CA, USA), with services exclusively provided by Biomics Biotech Co., Ltd. (Beijing, China). The raw sequencing data have been submitted to the NCBI’s Sequence Read Archive (SRA), with the accession number PRJNA1013739. To acquire clean reads, adaptors, poly-N sequences, and low-quality bases were eliminated. Subsequently, the HISAT2 tools software was employed to map the clean reads to the transcriptome sequence. Gene function annotation was carried out using multiple databases, namely Non-redundant (NR), euKaryotic Ortholog Groups (KOG), Clusters of Orthologous Groups (COG), Gene Ontology (GO), Swiss-Prot, Kyoto Encyclopedia of Genes and Genomes (KEGG), and Protein family (Pfam). The Fragments per Kilobase of transcript per Million fragments mapped (FPKM) values were used to represent gene expression levels.

The raw sequencing data have been deposited in the NCBI’s Sequence Read Archive (SRA; accession number: PRJNA1013739). Adaptors, poly-N sequences, and low-quality bases were removed to obtain clean reads. Gene function was annotated using the following databases: NR, KOG, COG, GO, Swiss-Prot, KEGG, and Pfam. Gene expression levels were represented by FPKM.

The seed samples used for iTRAQ analysis consisted of the same three biological replicates as those utilized in the RNA-seq experiment. Protein extraction, concentration, tryptic digestion, iTRAQ labeling, and LC-MS/MS analysis were performed according to the method described by Chen et al (Chen et al., 2022). Briefly, digested peptides were desalted using a Sep-Pak C18 column (Waters, Milford, MA, USA). The column was washed with 300 µL of Milli-Q water, and peptides were eluted with 50 µL of methanol. The eluate was dried by vacuum centrifugation and stored at –80 °C until further use. Peptide labeling was performed using an iTRAQ 4-plex kit (Applied Biosystems, USA) following the manufacturer’s protocol. Labeled peptides were also dried by vacuum centrifugation and stored at –80 °C prior to LC-MS/MS analysis.

LC-MS/MS analysis was carried out on a nanoElute system (plug-in V1.1.0.27; Bruker, Bremen, Germany) coupled to a timsTOF Pro mass spectrometer (Bruker, Bremen, Germany) equipped with a CaptiveSpray ion source. Raw data were processed using Peaks Studio X software (Bioinformatics Solutions Inc., Waterloo, ON, Canada). The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium with the dataset identifier PXD045963. To facilitate accurate protein identification, a species-specific protein database was constructed. This was achieved by first performing a de novo transcriptome assembly on RNA-seq data derived from the same seed batch, followed by predicting the coding sequences using TransDecoder (v5.5.0) with default parameters. All acquired MS/MS spectra were then queried against this custom database. A false discovery rate (FDR) threshold of ≤1% was applied, and the significance score [-10×log(p)] was calculated, where p denotes the probability that a peptide match occurred by chance.

The identification of significant differentially accumulated proteins (DAP) was filtered using thresholds of Q-value < 0.05 and |fold change (FC)| > 1.2, while DESeq2 software (based on a negative binomial distribution) was employed to analyze raw counts and identify differentially expressed genes (DEGs) from shared genes between all stages with the widely accepted criteria of |log₂ FC| ≥ 1 and adjusted P-value < 0.05. GO enrichment analysis was performed using the BiNGO plugin in Cytoscape, with an FDR cutoff of 0.05 for both DEGs and DAP. KEGG enrichment analysis was conducted using the KOBAS software for DEGs and DAP.

The quantification of endogenous phytohormones was conducted by MetWare (Wuhan, China). To measure the concentrations of phytohormones, including ABA, indole-3-acetic acid (IAA), jasmonic acid (JA), methyle jasmonate (MeJA), salicylic acid (SA), methyle salicylic acid (MeSA), gibberellin 1 (GA₁), gibberellin 3 (GA₃), gibberellin 4 (GA₄), gibberellin 6 (GA₆), gibberellin 7 (GA₇), gibberellin 8 (GA₈), gibberellin 9 (GA₉), gibberellin 12-ald (GA₁₂-ald), gibberellin 15 (GA₁₅), gibberellin 19 (GA₁₉), gibberellin 20 (GA₂₀), gibberellin 24 (GA₂₄), gibberellin 29 (GA₂₉), gibberellin 34 (GA₃₄), gibberellin 44 (GA₄₄), gibberellin 51 (GA₅₁), gibberellin 53 (GA₅₃), cis-zeatin-riboside (CZR), trans-zeatin-riboside (TZR), 6-(gamma,gamma-dimethylallylamino) purine (iP), 6-(gamma,gamma-dimethylallylamino) purine riboside (iPR), and 3-indolebutyric acid (IBA) were purchased from Sigma Aldrich (USA), approximately 1 g of seeds was ground in liquid nitrogen and extracted with acetonitrile.

The chromatographic separation was carried out on an ACQUITY UPLC CSH C18 column (2.1 × 100 mm, 1.7 µm; Agilent Technologies, Inc.). The mobile phase was composed of 0.05% (v/v) formic acid in a water–acetonitrile mixture (5:95, v/v). An injection volume of 10.0 µL was used, and the flow rate of the mobile phase was maintained at 0.35 mL/min. The column temperature was kept constant at 40 °C throughout the analysis. Phytohormones were detected and quantified in each sample using a UPLC-ESI-MS/MS system (UPLC, ExionLC™ AD, https://sciex.com.cn/, MS, QTRAP^® 6500+, https://sciex.com.cn/) after the addition of internal standards prior to grinding.

P. cyrtonema flowered from May to June and subsequently fruited from August to October. Fresh seeds extracted from mature capsules at the S1 stage had not yet achieved physiological maturity. At this stage, the endosperm occupied the majority of the seed volume, while the embryo showed no obvious differentiation, appearing rod-shaped and encapsulated by both the endosperm and the seed coat (Figure 1-left). After approximately 4 months of low-temperature storage, the endosperm near the seed pore began to degrade in the S2 stage (Figure 1-middle). During this stage, the embryo protruded from the seed pore, broke through the seed coat, and the seed commenced germination. As germination progressed, the embryo differentiated to form the primary rhizome and cotyledons during the S3 period, and the endosperm was fully absorbed (Figure 1-right). These three stages were critical in the process of embryo development and endosperm weakening. Investigating the differential genes, proteins, and metabolites during these periods might elucidate the molecular mechanisms underlying dormancy release and germination in P. cyrtonema seeds.

To investigate transcriptional dynamics during seed germination, nine samples representing three germination stages were subjected to RNA sequencing using the Illumina HiSeq platform. The RNA-Seq data are summarized in Table 1, Supplementary Table S1, and Supplementary Text S1, revealing the assembly of 197,609 unigenes with an N50 value of 270 bp. Functional annotation demonstrated that 775,597 (88.45%) unigenes showed homology to known sequences in the NR database, followed by 53,754 (61.27%) in KOG, 47,394 (54.02%) in COG, 43,817 (49.95%) in GO, 40,423 (46.08%) in Swiss-Prot, 40,112 (45.72%) in KEGG, and 30,132 (3.35%) in Pfam. Comparative genomic analysis revealed distinct sequence homology patterns: 19.35% of unigenes showed similarity to Asparagus officinalis, 6.40% to Fusarium sp., 5.68% to Rhizoctonia solani, with decreasing proportions for Elaeis guineensis (2.21%), Neonectria ditissima (2.18%), Phoenix dactylifera (1.63%), Ensete ventricosum (1.24%), Nothophytophthora sp. (1.16%), Fusarium oxysporum (0.78%), and Fusarium ambrosium (0.75%). Notably, 58.63% of sequences exhibited homology to other uncharacterized genomes (Supplementary Figure S1).

As shown in Supplementary Figure S2, the correlation coefficients of sample replicates were very high, indicating that the sequencing data were highly reliable. The number of unigenes identified at S1, S2, and S3 was 56,606, 71,097, and 54,818, respectively, with only 21,290 unigenes common to all three groups (Figure 2A). The number of DEGs between stages was as follows: 11,565 DEGs between S1 and S2, 10,196 between S1 and S3, and 7,696 between S2 and S3 (Supplementary Table S2). As illustrated in Figures 2B, C, the number of upregulated and downregulated DEGs varied across the comparisons. According to KEGG enrichment analysis, these DEGs were associated with 339, 340, and 321 pathways, respectively, of which 30 pathways were significantly enriched (Figure 2D). Among these pathways, those related to seed germination, phenylpropanoid biosynthesis, stilbenoid, diarylheptanoid, and gingerol biosynthesis, flavonoid biosynthesis, plant hormone signal transduction, and starch and sucrose metabolism exhibited the highest levels of enrichment. Therefore, the results suggest that plant hormone signal transduction pathways, as well as starch and sucrose metabolism pathways, may play significant roles in seed germination.

Seed proteome profiles were analyzed, and a total of 4,228 proteins were detected. The distribution of protein abundance across the three different stages of seed germination exhibited a nearly identical general pattern (Figure 3A; Supplementary Table S3). These identified proteins were further analyzed for GO enrichment. In the GO analysis (Figure 3B), “cell” was the most enriched subclass among cellular components, followed by “cell part,” “organelle,” and “membrane.” Within the category of molecular function, “catalytic activity” ranked highest, followed by “binding” and “localization.” In the biological process category, “metabolic process” was the most enriched, followed by “cellular process” and “biological regulation.” The results revealed that 647 proteins (449 increased and 198 decreased) were differentially expressed between S1 and S2, 1,720 proteins (1,044 increased and 676 decreased) between S1 and S3, and 1,475 proteins (848 increased and 627 decreased) between S2 and S3 (Figure 3C). KEGG annotation-based enrichment analysis revealed a set of significantly enriched protein functional terms across different pairwise comparisons. In the S1 vs. S2 comparison, “UDP-glucuronosyl and UDP-glucosyl transferase” was the most prominently enriched term, followed by “FOG: RRM domain”, “hydroxyindole-O-methyltransferase”, and “aspartyl protease”. For S1 vs. S3, “FOG: RRM domain” exhibited the highest enrichment, trailed by “aspartyl protease”, “serine/threonine protein kinase”, and “UDP-glucuronosyl and UDP-glucosyl transferase”. In the S2 vs. S3 comparison, “FOG: RRM domain” was again the most significantly enriched term, followed by “aspartyl protease”, “glyceraldehyde 3-phosphate dehydrogenase”, and “iron/ascorbate family oxidoreductases”. As illustrated in Figure 3D, several proteins—notably the FOG: RRM domain, aspartyl protease, and β-fructofuranosidase—were consistently and significantly enriched across all comparisons. This conserved pattern suggests that these proteins may play critical regulatory roles in seed germination.

To elucidate the regulatory relationship between transcriptional and translational processes, we conducted a systematic correlation analysis of proteomic and transcriptomic data. Comparative expression profiles of genes and their corresponding proteins across different experimental groups (S1, S2, S3) were presented in Figures 4A–C and Supplementary Table S4. The analysis revealed distinct patterns of expression coordination. A comparative analysis revealed distinct patterns of expression coordination among the three sample groups. The most significant translational changes were observed in the S2 vs. S3, accounting for 66.0% of all changes. Opposite changes were predominant in S1 vs. S2, making up 42.7% of the changes, whereas homodirectional changes were the main feature in S1 vs. S3 (14.2%). Notably, S2 vs. S3 exhibited the highest translational ratio and the lowest opposite/homodirectional ratio, indicating that translational changes played a predominant role in the late stage of seed germination.

KEGG pathway enrichment analysis of these differentially expressed modules identified key functional categories (Figure 4D). The concordant expression group showed predominant enrichment of aspartyl proteases (EC 3.4.23), β-fructofuranosidases (EC 3.2.1.26), and UDP-glycosyltransferases (EC 2.4.1.-). Conversely, the discordant expression category was dominated by molecular chaperones (particularly small heat-shock proteins Hsp26/Hsp42), followed by chitinase-related enzymes (EC 3.2.1.14). The translational regulation group exhibited significant enrichment of RNA-binding proteins containing FOG: RRM domains, along with β-fructofuranosidase (invertase) and arylacetamide deacetylase (EC 3.1.1.-) activities. Comprehensive pathway mapping identified three principal functional clusters associated with P. cyrtonema seed germination: (1) Genetic Information Processing (Ribosome, protein processing in the endoplasmic reticulum, spliceosome, RNA transport); (2) Metabolism (Starch and sucrose metabolism, cyanoamino acid metabolism, anthocyanin biosynthesis, phenylpropanoid biosynthesis, stilbenoid, diarylheptanoid, and gingerol biosynthesis, amino sugar and nucleotide sugar metabolism, ubiquinone and other terpenoid-quinone biosynthesis); (3) Signal transduction (Plant hormone signal transduction). These pathways were likely to play significant roles in seed germination.

During seed germination, 483 unigenes and 72 proteins were implicated in plant hormone signal transduction (Figure 5A; Supplementary Table S5). However, only 41 DAP were identified among them. Several phytohormone signal transduction-related genes exhibited consistent trends in mRNA and protein expression changes. For example, genes encoding transport inhibitor response 1 protein (TIR1) and GH3 auxin-responsive promoter (GH3) related to auxin, abscisic acid receptor PYL2 (PYL2), abscisic acid receptor PYR1 (PYR1) and gibberellin-regulated protein (GRP) related to ABA, and NADH oxidase (Nox) related to JA, all showed increased transcripts and protein abundance between S1 and S2 (Supplementary Table S5). In contrast, other genes/proteins between S2 and S3, such as those encoding nucleoside diphosphate kinase (NDK), PYR1, PYL2, 2-oxoglutarate-dependent dioxygenase (2OGD) and gibberellin receptor GID1 (GID1), exhibited opposite transcription levels (Supplementary Table S5). Overall, the upregulation and downregulation of these genes and proteins could lead to an imbalance of phytohormones during seed germination.

Dynamic profiles of ABA, IAA, JA, MeJA, SA, MeSA, GA₁, GA₃, GA₄, GA₆, GA₇, GA₈, GA₉, GA₁₂-ald, GA₁₅, GA₁₉, GA₂₀, GA₂₄, GA₂₉, GA₃₄, GA₄₄, GA₅₁, GA₅₃, CZR, TZR, iP, iPR, and IBA were analyzed across three developmental stages (S1, S2, S3) of P. cyrtonema seeds to track phytohormone dynamics (Figure 5B; Table 2). ABA, JA, and total GAs exhibited synchronized trends: levels decreased significantly from S1 to S2 but rebounded at S3. The contents of several bioactive GAs (He et al., 2019; Yamaguchi, 2008) changed significantly across stages. For instance, GA₁, GA₄, GA₉, and total bioactive GAs levels decreased, whereas GA₃, GA₆, and GA₇ levels increased from S1 to S2 or S3. Significantly, the bioactive GAs/ABA ratio increased from S1 to S3. Concurrently, most GA precursors, intermediates, and catabolites (e.g., GA₁₅, GA₂₄, GA₅₃, GA₃₄, etc.) also exhibited a decreasing trend from S1 to S2 or S3 (He et al., 2019; Yamaguchi, 2008). In contrast, IAA and SA contents declined progressively throughout all stages, with marked reductions from S1 to S3. CKs showed minimal fluctuations and remained at low concentrations, while IBA was undetectable in all samples.

Seed germination involved the dynamic rewiring of molecular networks, orchestrated by differentially abundant biomolecules. Figure 6 presented a proposed model of seed germination driven by endosperm weakening, which elucidated the germination mechanism of P. cyrtonema seeds (Supplementary Table S6). It showed that endosperm weakening was closely related to energy metabolism, carbohydrate metabolism and plant hormone signal transduction. Several key proteins in these pathways, such as beta-fructofuranosidase (INV), alpha-xylosidase (α-Xyl), beta-D-xylosidase (β-D-Xyl), and beta-glucosidase (bglX), promoted the conversion of starch into sucrose, and sucrose synthase (SUS) protein could promote the degradation of storage substances by the interaction of gibberellin and abscisic acid.