The experiments were conducted at the Industrial Crop Research Institute, Sichuan Academy of Agricultural Sciences, Qingbaijiang District, Chengdu City, Sichuan Province, China (30°42′N, 104°19′E), where average annual temperature was 18°C, and average annual precipitation was 558.1 mm (local weather bureau data). Salvia seedlings and maize, sesame, and soybean varieties were provided by the Institute of Economic Crop Breeding and Cultivation, Sichuan Academy of Agricultural Sciences. Basic Soil Properties: Potted Plants, pondus hydrogenii = 6.7, organic matter = 21.90 g/kg, available P = 43.76 mg/kg, available K = 115.50 mg/kg, total nitrogen = 1.48 g/kg; Field, potential of hydrogen = 6.0, organic matter = 19.78 g/kg, available P = 26.75 mg/kg, available K = 188.27 mg/kg, total nitrogen = 1.48 g/kg.

Pot experiments were set up in six planting combinations: maize + Salvia completely isolated (Salvia|Maize), maize + Salvia without isolation (SalviaMaize), soybean + Salvia completely isolated (Salvia|Soybean), soybean + Salvia without isolation (SalviaSoybean), sesame + Salvia completely isolated (Salvia|Sesame), and sesame + Salvia without isolation (SalviaSesame). S. miltiorrhiza was propagated from fresh root segments (1.3–1.8 cm in diameter) selected from seedlings dug in January 2022, and transplanted to pots and the field after seedlings were raised in the greenhouse. The pot experiment used 0.4-m-high, 0.6-m-diameter breeding bags, including two isolation treatments and was conducted in a randomized design ( Figure 1 ): ① plastic film isolation where the two crops were completely separated, with no inter-root intercropping; and ② without any isolation, where the two crops can communicate openly between the roots, with inter-root secretions interacting with each other, and where inter-root intercropping is obvious. Five replicates of each treatment were made.

The field experiments were conducted in a randomized design ( Figure 2 ), including soybean monoculture (soybeanm, 0.25 m spacing), maize monoculture (maizem, 0.4 m spacing), sesame monoculture (sesamem, 0.25 m spacing), Salvia monoculture (salviam, 0.25 m spacing), Salvia-maize intercrop (Salvia + maize), Salvia-soybean intercrop (Salvia + soybean), and Salvia-sesame intercrop (Salvia + sesame). The size of each experimental area was 3.6 × 5 m, and each treatment was replicated three times.

Soil samples were randomly taken from the field and pots at the time of sowing in March, and later mixed into composite samples that were naturally air-dried and stored in an oven for testing. Sampling was carried out in December 2022 on normal soil and inter-rhizosphere soil (soil close to plant roots and usually containing higher microbial populations than the surrounding normal soil) (Fierer, 2017). Inter-root soil (still attached to the roots) was shaken off gently and stored at −80°C for microbial community and inter-root metabolite analysis (Zhao et al., 2021). Simultaneously, soil around S. miltiorrhiza plants was collected, plant roots and stones were removed, and the soil was put into a self-sealing bag, and then it was dried and ground naturally, passed through a 60-mesh sieve and set aside for soil physical and chemical property testing.

Soil potential of hydrogen (pH) was determined by the point method (HJ 962-2018), where soil/water is mixed at 1/2.5 (w/v), stirred vigorously for 2 min, and left for 30 min, and then pH was measured with a pH meter (PHS-SE) (Pang et al., 2021). Soil organic matter (SOM) was determined using the Walkley–Black method, which oxidizes SOM by H₂SO₄ and K₂Cr₂O₇, followed by titration using FeSO₄ (Fu et al., 2015). Soil active phosphorus (available P) was measured by a spectrophotometer at 700 nm (NY/T 1121.7-2014) (Liu et al., 2018). Soil available potassium (available K) was determined by a flame photometer after extraction by 1 mol/L ammonium acetate (Saber et al., 1973). Soil total nitrogen was determined by the Kelvin digestion-constant distillation–titration method, followed by digestion using an automatic Kjeldahl nitrogen tester (Hennen K1100) (Wang et al., 2018).

A Soil DNA Kit was used to isolate soil microbial DNA from 5-g soil samples. The V4 hypervariable region of the bacterial gene was amplified with PCR using primers F (ACTCCTACGGGAGGCAGCA) and R (GGACTACHVGGGTWTCTAAT). After PCR processing, sequencing was conducted using an Illumina NovaSeq 6000. The raw image data files were converted to raw Sequenced Reads by Base Calling analysis. Sequenced data were filtered for connectors and low-quality using FASTP and Trimmomatic for Raw Reads to obtain high-quality valid sequences (Berg et al., 2015). Finally, the sequences were clustered with USEARCH (Edgar, 2013) (version 10.0) at the 97% similarity level, and OTUs were filtered using 0.005% of the total number of sequences sequenced as a threshold (Bokulich et al., 2013).

To analyze metabolites in inter-rhizosphere soil, a 50-mg sample was weighed and mixed with 1,000 μL of extraction solution containing internal standard. The mixture was then ground with steel beads and sonicated for 10 min in an ice water bath, followed by centrifugation, and the supernatant was carefully collected for testing. Metabolomics analysis was conducted using a liquid mass spectrometry system composed of a Waters Acquity I-Class PLUS UPLC tandem and a Waters Xevo G2-XS QTof high-resolution mass spectrometer (Acquity UPLC HSS T3, 1.8 μm 2.1*100 mm). Positive ionization mode: mobile phase A: 0.1% formic acid aqueous solution; mobile phase B: 0.1% formic acid acetonitrile; negative ionization mode: mobile phase A: 0.1% formic acid aqueous solution; mobile phase B: 0.1% formic acid acetonitrile. The Waters Xevo G2-XS QTof high-resolution mass spectrometer is capable of primary and secondary mass spectrometry data acquisition in MSe mode under acquisition software (MassLynx V4.2, Waters) control. In each data acquisition cycle, dual-channel data acquisition is possible for both low and high collision energies. Low collision energy was 2 V, the high collision energy range was 10–40 V, and scan frequency was 0.2 s for one mass spectral map. ESI ion source parameters were as follows: capillary voltage: 2,000 V (positive ion mode) or −1,500 V (negative ion mode); cone hole voltage: 30 V; ion source temperature: 150°C; desolvent gas temperature: 500°C; backblast gas flow rate: 50 L/h; desolvent gas flow rate: 800 L/h. Raw data collected by MassLynx V4.2 were processed by Progenesis QI software for peak extraction, alignment, and other data processing operations.

S. miltiorrhiza was harvested in December, its inter-root morphology was investigated, and its yield was measured. The number of branches, fresh weight, and dry weight of individual roots were also measured at harvest, along with root length (measured from the base of the stem) and diameter (the largest S. miltiorrhiza diameter) (Lei et al., 2018). After harvesting, the rhizome was dried, powdered, sieved (60 mesh), and stored to determine active ingredients. Intercrops were harvested in June and July.

S. miltiorrhiza active ingredients (tanshinones and tannins) were determined using high-performance liquid chromatography. Total tanshinones and tannic acid B were determined according to the Chinese Pharmacopoeia 2015 edition, wherein tanshinones (Tanshinone IIA, Cryptotanshinone, and Tanshinone I) shall not be<0.25% in total and tanshinolic acid B shall not be<3.0%. Chromatographic analysis was performed with high-performance liquid chromatography equipment (Agilent 1260 series, USA) on a ZORBAXSB-C18 (4:6 mm × 250 mm, 5 μm) column at a temperature of 20°C. For salvianolic acid B determination, the mobile phase consisted of acetonitrile (A) and 0.1% phosphoric acid aqueous solution (B) in gradient elution mode for 20 min at 90% A with a flow rate of 1.0 mL min⁻¹ and an injection volume of 10 μL at a detection wavelength of 286 nm. The mobile phase was composed of acetonitrile (A) and 0.02% phosphoric acid aqueous solution (B): 0–6 min, 0%–61% A; 6–20 min, 61%–90% A; 20–20.5 min, 90%–61% A; 20.5–25 min, 61% A, with a flow rate of 0.8 mL min⁻¹, a sample volume of 10 μL, and a detection wavelength of 270 nm (Liu et al., 2020a).

To measure land area efficiency when intercropping, Willey and Osiru (1972) introduced the land equivalent ratio (LER) concept, which reflects the yield of two crops as intercrops and as a monocrop, which can be used to evaluate the suitability of two crops for intercrop cultivation. It is:

where p LER is a partial land equivalent ratio, Y₁ and Y 2 are the yields of seed 1 and seed 2 in interplanting, M₁ and M 2 are the yields of seed 1 and seed 2 in single crop, respectively.

All statistical analyses were performed using various packages in R version 3.3.0 (Vermeulen et al., 2013), including ANOVA tests for multiple comparisons of soil microbial diversity, soil properties, active ingredient content, and plant biomass. Root secretion aroma diversity and evenness indices were compared among the groups by unpaired t-tests. When p< 0.05, the means of the groups were considered significantly different using the least significant difference (LSD). Principal component analysis (PCA), redundancy analysis (RDA), and heatmaps were generated using the “ggplot2” package in the R platform. PCA regroups all the metabolites originally identified linearly to form a new set of variables to determine differences between the groups of samples. To determine the proportion of variation in community structure explained by environmental factors, microbial community compositional components were determined by RDA, followed by variance decomposition analysis (VPA). Heatmaps were used to compare correlations between species abundance (OTUs) and various indicators, including environmental factors, plant growth parameters, and root base secretions.

In commercial trade, S. miltiorrhiza yield, quality, and appearance are very important for Salvia growers to achieve profitability. The S. miltiorrhiza fresh weight index can represent its yield, while hydrophilic and lipophilic components as well as root length and diameter indicate its quality and appearance (Liu et al., 2020a). Inter-root secretions between different plants had a significant effect on Salvia morphology ( Table 1 ), and Salvia dry weight in the SalviaSesame treatment increased by 3.21% compared to that in Salvia|Sesame when the plants were compared with and without the isolated treatment. S. miltiorrhiza fresh weight, dry weight, and root diameter were significantly decreased by inter-root secretions from maize, and all Salvia growth indices were decreased in SalviaSoybean and SalviaMaize compared to SalviaSesame. Overall, inter-root secretions from sesame had the least effect on S. miltiorrhiza root development. The combinations SalviaMaize and SalviaMaize significantly inhibited growth, as Salvia fresh weight, dry weight, and root diameter decreased by 46.29%, 46.77%, and 27.88%, respectively.

Active ingredient contents were compared under the six cultivation combinations ( Figure 3 ). There were no significant differences in tanshinolic acid B. After SalviaSesame treatment, S. miltiorrhiza tanshinones significantly increased, with the highest levels of Tanshinone IIA, Tanshinone I, and Cryptotanshinone, showing increases of 27.06%, 22.76%, and 26.41%, respectively, compared to Salvia|Sesame. Sesame and Salvia cultivation effectively increases accumulation of Tanshinone active ingredients.

In the potted soil after each treatment group (Salvia|Soybean, SalviaSoybean, Salvia|Maize, SalviaMaize, Salvia|Sesame, and SalviaSesame), there were no significant differences in acidity ( Figure 4A ), available P ( Figure 4C ), organic matter ( Figure 4D ), and total nitrogen ( Figure 4E ). The highest available K ( Figure 4B ) was found in SalviaSoybean (152.13 ± 13.69), which was significantly higher than available K in SalviaSesame (122.94 ± 15.96), while there were no significant differences between plants with and without compartment treatments.

Illumina detected a total of 1,121,526 clean reads generated from the 16S RNA gene in 18 samples, and a total of 8,654 OTUs were identified in all samples. The Chao1 index (a measure of α-diversity) was significantly lower for SalviaSoybean and SalviaMaize than for SalviaSesame, with SalviaSoybean having the lowest (275.67 ± 83.28c) number of species ( Table 2 ). The Shannon and Simpson indices indicated that bacterial diversity and homogeneity were typically better for all treatments, with greater variation for SalviaSoybean.

Additionally, there were differences in bacterial community structure after the treatment in each group. The highest abundance at bacterial phylum level was found in Proteobacteria (26.41%–44.04%), Acidobacteriota (8.37%–26.88%), Gemmatimonadota (4.62%–17.40%), Actinobacteriota (3.78%–9.13%), and unclassified_Bacteria (2.28%–11.44%) ( Figure 5A ). Cyanobacteria and Fusobacteriota significantly increased in the SalviaSoybean treatment, accounting for 29.11% and 4.34% of total community abundance, respectively, while both did not exceed 0.1% in any other treatment. SalviaMaize and SalviaSoybean bacterial community abundance and number showed a significant decrease compared to Salvia|Maize and Salvia|Soybean, while SalviaSesame and Salvia|Sesame bacterial community structure did not. There was no significant difference in the structure of SalviaSesame and Salvia|Sesame bacterial communities. An RDA was performed to accurately assess the relative effects of soil physicochemical properties (potential of hydrogen, organic matter, total nitrogen, available P, and available K) on the abundance and diversity of S. miltiorrhiza inter-rhizosphere bacterial communities. The first two RDA components explained 51.48% and 19.94% of the overall variation, respectively. Furthermore, potential of hydrogen, available P, available nitrogen, total nitrogen, and organic matter appeared to play a crucial role in building bacterial communities ( Figure 5B ). There was a negative correlation between organic matter and RDA1 (70.64%), a direct correlation between available P and RDA1 (59.13%), and a maximum contribution of 50.81% from total nitrogen to RDA2. This indicates that organic matter, available P, and total nitrogen remain the key environmental factors affecting Salvia inter-root microorganism abundance. The soil environmental factor and microbial communities correlation analysis revealed that the potential of hydrogen and Gemmatimonadota were significantly negatively correlated (r ² = 0.74; p< 0.001), and available K and Patescibacteria were significantly negatively correlated (r ² = 0.74; p< 0.001) ( Figure 5C ).

After analyzing the inter-root microbial community and agronomic trait indicators of Salvia divinorum, we found that sesame was the most promising plant to grow with it. The relationship between sesame and S. miltiorrhiza inter-root secretions was further analyzed by PCA, and there were similar metabolites between SalviaSesame and Salvia|Sesame ( Figure 6A ). The Salvia divinorum inter-root microbial community was analyzed when the three crops were not isolated, and the lowest microbial community α-diversity was under SalviaSoybean treatment, while the highest microorganism abundance and number were found under SalviaSesame treatment. This may be due to different crops resulting in different inter-root S. miltiorrhiza feedbacks. Analysis of SalviaSoybean and SalviaSesame inter-root secretions revealed that 30 were significantly reduced in the SalviaSoybean treatment (p< 0.05 and VIP > 1.5) ( Figure 6B ), of which 18 belonged to the Human Metabolome Database (HMDB), with lipids and lipid-like compounds, nucleosides and analogs, organic acids and derivatives, organic oxides, and heterocyclic compounds accounting for 4.05%, 26.56%, 2.84%, 15.20%, and 8.27%, respectively. The two lowest p-value secretions screened, D-Sedoheptulose 7-phosphate and Nicotinamide-beta-riboside, both belong to the organic oxide class, and comparing the amounts of these two metabolites in SalviaSoybean and SalviaSesame, the difference multiplicity (fold change) was −1.63746 and -2.02996 ( Figure 6C ), respectively. The correlation heatmap between S. miltiorrhiza yield, quality, and active ingredients showed that lipids were significantly and negatively correlated with yield (fresh weight and dry weight) and quality (Tanshinone IIA, Tanshinone I, and Cryptotanshinone) ( Figure 7 ).

To accurately assess the effects of microbial communities on root secretions and S. miltiorrhiza growth and development, we investigated the relationships between the main secretions (lipids, nucleosides and analogs, organic acids and derivatives, organic oxygen compounds, organoheterocyclic compounds, phenylpropanoids, and polyketides), Salvia basic growth and developmental indicators (fresh weight, dry weight, and number of nucleosides), Salvia active constituents (Salvianolic acid B, Tanshinone IIA, Tanshinone I, and Cryptotanshinone), and the main microbial communities. Lipids and lipid-like molecules and Bacteroidota, Chloroflexi, Nitrospirota, Planctomycetota, and Proteobacteria were significantly negatively correlated (p< 0.05) ( Figure 8 ). Organic nitrogen compounds, root width, dry weight, and Patescibacteria all showed significant positive correlations (p< 0.05). Fresh weight, dry weight and Bacteroidota, Planctomycetota were significantly positively correlated with each other (p< 0.05). Interestingly, Bacteroidota, Planctomycetota, and Patescibacteria were all significantly positively correlated with each other (p< 0.05).

Compound planting of different plants and Salvia led to differences in harvest yield and LERs. Averaging the data from three replications, S. miltiorrhiza yield after intercropping maize, soybean, and sesame was 9,817.20 ± 322.20, 14,817.451,950.15, and 15,739.65 ± 3,061.20 kg/hm², respectively, while monocropping maize, soybean, sesame, and Salvia yields were 17,385.001,967.7, 3,432.45 ± 488.85, 1,800.15361.20, and 16,470.00 ± 3,616.80 kg/hm², respectively. The lowest Salvia yield was after intercropping with maize, where it was reduced by 40.44%. Salvia yield was reduced by 10.11% and 4.52% when intercropped with soybean and sesame, respectively. The highest LER (1.82) was achieved with sesame and Salvia intercropping, with pLER₁ and p L E R 2 of 0.96 and 0.86, respectively, and the lowest LER of 1.36 for soybean and Salvia intercropping. Field experiments verified that sesame and Salvia are a good intercropping combination.