The experiment was conducted at the main fruit tree germplasm resource garden of the Gansu Academy of Agricultural Sciences (N 36°6’, W 103°41’, altitude 1530 m). This region is characterized by an average annual rainfall of 329 mm, an average annual temperature of 9.6°C, an extreme minimum temperature of -25°C, and a frost-free period of 196 days. The experimental soil was classified as loessial, with the preceding vegetation consisted of three-year-old dandelions in a fallow orchard plot. Soil analysis of the topsoil layer (0-20 cm) revealed the following properties: organic matter content, 22.3 g/kg; total nitrogen (TN), 1.45 g/kg; total phosphorus (TP), 1.18 g/kg; total potassium (TK), 22.06 g/kg; available nitrogen (AN), 117.71 mg/kg; available phosphorus (AP), 28.90 mg/kg; available potassium (AK), 334.30 mg/kg; and pH, 8.37. Prior to the experiment, topsoil samples were collected and sieved to remove debris for subsequent analysis. One-year-old grafted walnut seedlings of the “Yuanlin” cultivar—grafted onto rootstocks derived from native late-fruiting thick-shelled walnuts in Cheng County, Longnan City, Gansu Province—were transplanted on March 25, 2019. The soybean cultivar “Longhuang 3”, a locally dominant variety, which was sown on April 15, 2020. Both cultivars are widely cultivated in the study region. Two planting systems were established: walnut monoculture and walnut-soybean intercropping (Figure 1). The experiment was conducted using planting bags (140 cm × 140 cm, 1.96 m²) as individual experimental units (plots). The walnut trees had a main trunk height of 60 cm and a crown width of 80 cm. In the walnut-soybean intercropping system, soybeans were sown in a single north-south row on both the east and west sides, with a row-to-trunk distance of 50 cm. Intra-row spacing of soybean plants was 20 cm, resulting in 10 plants per planting bag. A completely randomized design was adopted, with nine biological replicates per treatment. All experimental units were subjected to local conventional agricultural management practices (e.g., irrigation, weeding) without additional nitrogen fertilization.

Sampling was conducted at three critical stages: the hard kernel stage of fruit development (June 25, 2020), the oil synthesis period (August 15), and before walnut leaf abscission (October 5). Destructive sampling of whole plant was performed, with walnuts divided into four components (main trunk, branches, leaves, and root system) and soybeans divided into three components (pods, straw, and root system). After transporting the samples to the laboratory, the fresh weight of each component was measured. Subsequently, the samples were blanched at 105°C for 30 minutes and then dried at 80°C until a constant weight was achieved. Following cooling, the dry weight of each component was recorded. A portion of the dried samples was ground, passed through a 0.25 mm sieve, and homogenized for storage. The total nitrogen content of the plant samples was determined using the Kjeldahl method. Soil samples collected from the walnut rhizosphere were promptly transported to the laboratory for air-drying, grinding, and sieving before subsequent analyses. The TN, total phosphorus (TP), and total potassium (TK) contents of the soil were determined using the automatic nitrogen determination method (Li et al., 2016), the molybdenum-antimony colorimetric method (Wang et al., 2014), and the sodium hydroxide fusion method (Liu et al, 2023), respectively. Soil organic matter (SOM) content was determined via the potassium dichromate oxidation method (Liu et al., 2023). Soil available nitrogen (AN), available phosphorus (AP), and available potassium (AK) were determined using the alkali diffusion method (Razgulin, 2009), the sodium bicarbonate extraction followed by molybdenum-antimony-ascorbic acid colorimetric method (Mussa et al., 2009), and the neutral ammonium acetate extraction method (Wang et al., 2014), respectively.

The metabolite extraction was conducted using the organic reagent-based protein precipitation method, and quality control (QC) samples were prepared simultaneously. The QC samples were generated by pooling equal volumes of the prepared experimental samples. The extracted samples were randomized in sequence, with QC samples interspersed before, during, and after the sample set to evaluate the reproducibility of the experimental technique (Dunn et al., 2011). Mass spectrometry analysis was performed in both positive and negative ion modes. Raw data acquired from the TripleTOF5600plus mass spectrometer (SCIEX, Framingham, MA) were converted into a readable mzXML format using the MSConvert software from ProteoWizard. Peak detection and alignment were carried out using XCMS software, followed by quality control of the peak extraction process. Adduct ions of the detected compounds were annotated using CAMERA, and preliminary identification was conducted using the metaX software (Wen et al., 2017). During the identification process, both primary and secondary mass spectrometry data were matched against a self-constructed standard compound database. Candidate metabolites were further annotated using public databases such as Human Metabolome Database (HMDB 5.0: https://hmdb.ca/) and Kyoto Encyclopedia of Genes and Genomes (KEGG: https://www.kegg.jp) to elucidate their physicochemical properties and biological functions. Differential metabolites were quantified and screened using the metaX software.

In this study, six samples were sequenced on the Illumina NovaSeq 6000 sequencing platform (Illumina, San Diego, CA), and the raw sequencing data were analyzed by Shenzhen Huada Gene Technology Service Co., Ltd. Initially, total RNA was extracted from the samples using a plant RNA extraction kit (R0017M, Beyotime), and its concentration, purity, and integrity were evaluated using a NanoDrop spectrophotometer, Qubit^® 2.0 fluorometer, and an Agilent 5400 bioanalyzer, respectively. Subsequently, mRNA was enriched from total RNA, fragmented, and reverse-transcribed into cDNA using random hexamer (N₆) primers. After second strand cDNA synthesis, the double-stranded DNA was subjected to end repairs, phosphorylation, and 3’-A overhang addition, followed by ligation with bubble adapters. The ligation products were amplified by PCR, denatured to single-stranded DNA, and circularized to construct libraries for DNBSEQ sequencing.

The raw data generated from sequencing are referred to as raw reads. To ensure data quality, low-quality reads, adapter contamination, and reads with excessive unknown base ‘N’ content were filtered out, resulting in clean reads. The clean reads were aligned to the reference genome for new transcript prediction, SNP and InDel detection, and alternative splicing gene analysis. Newly identified transcripts with protein-coding potential were integrated into the reference gene sequences to form a comprehensive reference sequence set. Gene expression levels were subsequently calculated. Differentially expressed genes (DEGs) were filtered using thresholds of |log2 (fold change)| ≥ 1 and false discovery rate (FDR) < 0.05. Finally, DEGs between different samples were detected as required, and in-depth clustering and functional enrichment analyses were conducted. Pearson correlation coefficients were used to assess the biological variability of DEG expression among samples, where an R² value closer to 1 indicates a stronger correlation between samples.

To validate the reliability of the root transcriptome sequencing results, qRT-PCR was performed on six genes associated with nitrogen metabolism in the roots. First-strand cDNA synthesis was carried out using the HiScript II 1st Strand cDNA Synthesis Kit (R212-01, Vazyme Biotech). Subsequently, qRT-PCR was executed on the Bio-Rad CFX96^™ Real-Time PCR Detection System (Bio-Rad), employing ChamQ SYBR Color qPCR Master Mix (Q411-02, Vazyme Biotech). The relative expression levels of the target genes were calculated using the 2^-ΔΔCt method. The primers used for qRT-PCR analysis are listed in Supplementary Table S1.

All experimental data were statistically analyzed and subjected to one-way analysis of variance (ANOVA) using SPSS 22.0 software. For multiple comparisons following ANOVA, Tukey’s honestly significant difference (HSD) test was used to determine statistical significance (P < 0.05). Graphs were generated using GraphPad Prism 8.3 software.

From the observation of root system morphology (Figure 2A), significant differences in walnut root architecture were identified under different planting patterns. Specifically, the vertical roots of intercropped walnut progressively extended downward during growth and exhibited more pronounced development compared to horizontal roots. In contrast, the root distribution of monocropped walnut was relatively uniform. To accurately quantify these morphological differences, further determinations of aboveground and root dry matter accumulation were conducted.

Under both planting patterns, the dry matter mass of both aboveground and root components consistently increased, peaking in October. Notably, starting from August, the rate of dry matter accumulation accelerated significantly (Figure 2B). Compared with June, by October, the aboveground dry matter accumulation of intercropped and monocropped walnuts increased by 382.16% and 309.11%, respectively, while the root dry matter mass increased by 505.41% and 215.75%, respectively. It is worth noting that the total dry matter accumulation of intercropped walnut was higher than that of monoculture walnuts, indicating that intercropping has a significant advantage in promoting dry matter accumulation.

In cereal-legume intercropping systems, leguminous crops can reduce reliance on chemical nitrogen fertilizers through biological nitrogen fixation and transfer a portion of nitrogen to neighboring cereal crops. This provides an additional nitrogen source for the growth and development of cereal crops (Kebede, 2021; Landschoot et al., 2024). However, the nitrogen allocation in walnut root within walnut-soybean intercropping systems remains poorly understood and requires further investigation. Our results showed that during the hard kernel period and nut maturity period, the nitrogen content in the aboveground component of intercropped walnuts with soybeans was lower than that in monoculture. This phenomenon may be related to potential interspecific interactions: the nitrogen-fixing capacity of soybeans (a key characteristic in such intercropping systems) and the likely prioritization of nutrient allocation to developing walnut fruits could collectively contribute to the observed difference, though this requires further verification through targeted nutrient allocation studies. Conversely, during the dormant period, the nitrogen content in the aboveground parts of walnuts in intercropping is higher than in monoculture, which may be related to the nutrient retranslocation to the tree body and the continuous absorption by the root system. Similarly, the nitrogen content in the roots of walnut trees exhibits a comparable trend, being lower than in monoculture during the nut maturity period but higher during the hard kernel period and dormant period (Figure 3A).

Intercropping leguminous plants in orchards increases soil organic carbon, microbial carbon, and total nitrogen contents. Compared with monoculture, intercropping increased SOM content during the hard kernel stage and nut maturity stage but decreases it during the leaf-falling stage, ensuring a continuous nutrient supply for walnut growth (Figure 3B). During the hard kernel stage and nut maturity stage, intercropping reduces TN, TP, AN, AP, and AK contents in the soil compared to monoculture. This observation reflects intensified soil nutrient utilization in the intercropping system. During the dormant period, as walnut nutrient absorption decreased, soil nutrients are able to recover. Furthermore, after soybean harvest, soil nutrient contents in both intercropping and monoculture systems tend to converge (Figure 3B).

Metabolites extracted from the samples were analyzed using a high-resolution mass spectrometer in both positive and negative ion modes, generating comprehensive mass spectrometry datasets. The detailed statistical results of metabolite detection are presented in Supplementary Table S2. To ensure data quality, the total ion intensities of all detected metabolites in each sample were summed and quality validation was performed using total ion chromatograms (Supplementary Figure S1). To visualize the differences among sample groups, principal component analysis (PCA) and partial least squares-discriminant analysis (PLS-DA) were conducted on 16 samples in both positive and negative ion modes (Supplementary Figure S2). The PCA results showed a distinct separation between the two treatment groups along the first principal component (PC1), which accounted for 44.34% of the total variance, while PC2 explained an additional 20.19% of the variation. These results support the consistency and reliability of the analytical procedure. PLS-DA analysis further confirmed a clear separation between the intercropping and monoculture systems, highlighting the metabolic divergence induced by different planting patterns.

Figure 4A illustrates the number of differential metabolites identified in both positive and negative ion modes, clearly demonstrating significant metabolic differences between walnut intercropping and monoculture systems. Specifically, 833 and 556 metabolite ion features were closely associated with the two planting patterns under positive and negative ion modes, respectively. Notably, the number of upregulated metabolites was greater than that of downregulated ones under intercropping conditions compared to monoculture, suggesting that intercropping not only induces metabolic variation but also promotes the accumulation of specific metabolites. Furthermore, Figure 4B presents a hierarchical clustering heatmap, revealing a distinct separation of average metabolite profiles across the different experimental groups, thereby validating the observed metabolic distinctions.

To examine the quantitative changes and identify differential metabolites, primary and secondary identification annotations were performed, and the secondary-identified differential metabolites were visualized (Figure 5A; Supplementary Table S4). The results demonstrated that intercropping with soybeans significantly altered the metabolite profiles in the walnut root system. A total of 80 differential metabolites were identified, primarily including lipids, nucleotides, phenylpropanoids, organic heterocycles, organic acids, and phenyl derivatives. Notably, the upregulated metabolites (22 in total) were predominantly phenylpropanoids, organic nitrogen compounds, organic acids, and lipids, while the downregulated metabolites (58 in total) were predominantly nucleotides, phenyl derivatives, and most lipids were significantly downregulated.

Subsequently, based on pathway classifications from the KEGG database, we conducted enrichment analysis on the secondary-identified differential metabolites (Figure 5B). All enriched metabolites were categorized into six major classes: metabolism, genetic information processing, environmental information processing, cellular processes, human diseases, and organismal systems. Among these, metabolic pathways exhibited the highest enrichment degree. The top seven enriched metabolic pathways included global and overview maps, lipid metabolism, amino acid metabolism, carbohydrate metabolism, other amino acid metabolism, nucleotide metabolism, and cofactor and vitamin metabolism. These pathways play fundamental roles in regulating cell proliferation, differentiation, and overall plant growth during walnut root development. Additionally, the pathway related to the biosynthesis of secondary metabolites and energy metabolism—ranked among the top enriched pathways, which is closely associated with the allelopathic effects of secondary metabolites, such as paeonol, secreted by walnut roots. KEGG enrichment analysis revealed that under intercropping conditions, pathways related to carbon and nitrogen metabolism, including amino acid metabolism, carbohydrate metabolism, and nucleotide metabolism, were significantly enriched. Collectively, these findings indicate that intercropping induces substantial alterations in carbon and nitrogen-related metabolic pathways within the walnut root system.

A total of 259.46 million raw reads were generated through sequencing in this study. After rigorous quality filtering, the Q20 values (GC content-related) all surpassed 99.69%, while the Q30 values exceeded 92.03%, and the clean read ratio was above 96.47% (Supplementary Table S3). These metrics meet the stringent quality control standards for sequencing data. The Pearson correlation coefficients among the three biological replicates were high (Supplementary Figure S3), indicating robust data consistency and reliability, which to supports further analysis. The percentage of reads successfully mapped to the reference genome ranged from 87.13% to 93.15%, with the proportion of uniquely mapped reads falling between 58.89% and 62.98% (Supplementary Table S5).

When comparing DEGs between walnut roots in the intercropping and monoculture systems, a total of 3,978 DEGs were identified, including 2,452 upregulated and 1,526 downregulated genes (Figure 6A). To elucidate the functional implications of these DEGs, Gene Ontology (GO) enrichment analysis was conducted on walnut roots under both planting conditions. Through this analysis, the DEGs were categorized into three main classes: biological processes, molecular functions, and cellular components. Notably, the term “nitrogen utilization” was significantly enriched, which is consistent with the findings from metabolomics analysis (Figure 6B).

The KEGG integrates genomic, chemical, and system-level functional information comprehensively. In this study, KEGG pathway enrichment analysis was performed to explore metabolic pathways in walnut root systems under different planting patterns. Following classification by pathway type, metabolic pathways exhibited the highest degree of enrichment. Consistent with the results of metabolomics analysis, several key pathways were significantly enriched, including global and overview maps, lipid metabolism, carbohydrate metabolism, amino acid metabolism, terpenoid and polyketide metabolism, as well as biosynthesis of other secondary metabolites. Additionally, pathways related to carbon and nitrogen metabolism, such as amino acid metabolism, sugar metabolism, and nucleic acid metabolism, were also significantly enriched (Figure 6C). Based on the integrated metabolomics and transcriptomics results, it can be concluded that walnut root systems under intercropping conditions undergo substantial alterations in carbon and nitrogen-related metabolic pathways.

DEGs and differential metabolites were separately mapped to the KEGG pathway database to identify commonly enriched pathways shared between the two datasets, thereby elucidating the key biochemical and signal transduction pathways potentially involved in their regulatory mechanisms. A total of 32 co-enriched KEGG pathways were detected in both intercropping and monoculture systems (Figure 7A), with nitrogen metabolism being significantly enriched. Moreover, several pathways closely associated with carbon and nitrogen metabolism, including amino acid metabolism, sugar metabolism, and nucleic acid metabolism, also showed significant enrichment.

Supplementary Table S6 provides a comprehensive list of differentially expressed genes related to nitrogen metabolism identified through GO enrichment analysis. Figure 7B illustrates the nitrogen metabolism pathway along with the heatmap representing the expression patterns of some genes. Key enzymes and transporters, including NR, NIR, GOGAT, GDH, GS, nitrate transporter (NRT), and AMT play essential roles in nitrogen metabolism. The expression trends of these genes in the intercropped walnut root system varied, reflecting their distinct functions and contributions to the nitrogen metabolic pathway. To validate the RNA-seq gene expression data presented in the heatmap, we conducted qRT-PCR for further confirmation. The results demonstrated that the expression levels of several critical genes in the walnut root system were significantly altered under intercropping compared to monoculture. Specifically, genes such as 108988099 (GS), 109005551 (NRT), 109004582 (GOGAT), 108994933 (AMT), and 108982946 (NIR) exhibited upregulation ranging from approximately 1.77-6.65 fold. Notably, the expression level of the 109012439 (NR) gene decreased after intercropping (Figure 7C). This downregulation may be attributed to feedback inhibition caused by increased NR enzyme activity in the walnut-soybean intercropping system. Consistently, the contents of several intermediate products involved in the nitrogen metabolism pathway, such as L-glutamine, glutamine, D-gluconic acid, uridine, L-tryptophan, phytosphingosine, and guanine, were markedly elevated in the walnut root intercropping system compared to the monoculture (Figure 7D). These findings suggest that nitrogen metabolism intermediates may play a crucial role in the walnut intercropping system.

As illustrated in Figure 8, under intercropping conditions, the enzymatic activities associated with nitrogen metabolism in walnut root systems exhibited varying degrees of change. The activities of the four analyzed enzymes were consistently higher in the intercropping treatment compared to the monoculture treatment. Specifically, significant differences in the activities of NR, GS, and GOGAT were observed between the intercropping and monoculture treatments in August, whereas no significant differences were noted in October. In contrast, NIR demonstrated significant differences between the two treatments even in October. These findings suggest that intercropping with soybeans enhances the GS/GOGAT cycle pathway, which is responsible for converting inorganic nitrogen to organic nitrogen in walnut root systems. Furthermore, these results consistent with the changes in nitrogen-related metabolic pathways identified through metabolomics analysis.