The experiment was conducted in 2024 in Saihan District, Hohhot, Inner Mongolia Autonomous Region, China (40°48′26″ N, 111°44′48″ E). The region has a temperate continental monsoon climate, with an average annual temperature ranging from 6.3 °C to 7.7 °C and mean annual precipitation of 138.2 mm. Annual evaporation varies from 2,030 mm to 2,700 mm, and the frost-free period lasts approximately 133–144 days. The soil type was classified as chestnut soil. To document the pre-treatment (baseline) conditions shared by all pots, soil properties were measured prior to salinity application: electrical conductivity (EC) 0.27 mS cm^-¹, pH 7.89, nitrate nitrogen (NO₃^-–N) 1.02 mg kg^-¹, total nitrogen (TN) 980.48 mg kg^-¹, and total phosphorus (TP) 550.09 mg kg^-¹.

A pot experiment was performed using alfalfa (Medicago sativa L.) cultivar ‘Algonquin’. Field-collected soil was passed through a 4-mm sieve, homogenized, and air-dried under shade prior to use. Alfalfa seeds were germinated in plug trays, and seedlings were transplanted into pots four weeks after emergence. Each plastic pot (26.6 cm diameter, 24.5 cm height) was filled with 7.5 kg of sieved soil. After thorough irrigation, 20 seedlings were transplanted into each pot.

Salinity was imposed by mixing a neutral salt blend of NaCl and Na₂SO₄ (1:1, w/w) into the soil to achieve two target levels: low salinity (LS, 3.0 g kg^-¹ dry soil) and moderate salinity (MS, 4.5 g kg^-¹ dry soil), with a non-saline control (CK). The NaCl + Na₂SO₄ mixture was used because chloride and sulfate salts commonly co-occur in saline soils, making mixed neutral salts more representative than NaCl alone (Reginato et al., 2021). The two application rates (3.0 and 4.5 g kg^-¹) were selected to impose mild and moderate salinity stress while avoiding excessively severe conditions frequently used in pot studies (Hou et al., 2022; Irakoze et al., 2021). Salts were thoroughly mixed into the soil once at the start of the treatment, after which all pots received the same irrigation regime. All pots were irrigated with 500 mL of deionized water per day; the control (CK) received the same water volume as the salinity treatments. Salinity treatments started on 17 June 2024 (at transplanting stage) and were maintained until harvest on 30 September 2024 (105 days; Figure 1).

At harvest, both plant and soil samples were collected. Surface debris, litter, and gravel were removed prior to sampling. In this study, “root-zone soil” refers to soil collected adjacent to the visible root system within each pot. Root-zone soil was obtained by coring soil within approximately 3 cm of the roots. We did not physically separate a strictly defined rhizosphere fraction (i.e., soil tightly adhering to the root surface recovered by washing); therefore, the microbial profiles represent communities from the operationally defined root-zone soil compartment rather than strictly defined rhizosphere soil. Soil from each pot was divided into two subsamples: one was placed into 25-mL centrifuge tubes, preserved on dry ice, and transported for metagenomic sequencing; the other was sealed in plastic bags for measurements of soil pH, electrical conductivity (EC), NO₃^--N, NH₄⁺-N, and total nitrogen (TN).

All statistical analyses were performed in R (v4.4.3). Data were assessed for normality using the Shapiro–Wilk test. For soil physicochemical properties and plant growth traits, differences among treatments (CK, LS, and MS) were evaluated using one-way ANOVA followed by Tukey’s HSD for post hoc comparisons. For metagenome-derived relative abundance data (e.g., dominant genera and functional categories/pathways), Kruskal–Wallis tests were applied, followed by Dunn’s post hoc tests for pairwise comparisons. P values from multiple comparisons were adjusted using the Benjamini–Hochberg false discovery rate (FDR) procedure. Unless stated otherwise, statistical significance was accepted at FDR-adjusted p < 0.05 after FDR correction, or at p < 0.05 for single tests. Microbial α-diversity indices (Shannon and Simpson) were compared among treatments using one-way ANOVA followed by Tukey’s HSD (as shown in Figure 2). Microbial β-diversity was assessed using principal coordinates analysis (PCoA) based on Bray–Curtis dissimilarities, and treatment effects on community composition were tested using PERMANOVA (e.g., adonis2). PERMANOVA results are reported with effect sizes (R²) and permutation-based P values. Given the low biological replication (n = 3 per treatment), statistical power—particularly for multivariate and metagenome-inferred functional comparisons—is limited; therefore, ordination patterns and functional results are interpreted cautiously, and trend-level patterns are explicitly identified where appropriate. Plots were generated in R using ggplot2 and related packages. Heatmaps and additional visualizations were rendered using the Lingbo MicroClass visualization platform.

As shown in Figure 3, soil physicochemical properties at harvest differed among treatments. Soil EC increased under salinity relative to the control (approximately 290 μS cm^-¹ in CK vs. 480–510 μS cm^-¹ under salinity), with LS showing higher EC than MS (p < 0.05). Soil pH increased along the salinity gradient (CK< LS < MS), and all pairwise differences were significant (p < 0.05).

Salinity also altered soil nitrogen status. TN was higher under salinity than in the control (p < 0.05). NO₃^--N increased under moderate salinity compared with CK and LS (p < 0.05), whereas CK and LS did not differ. In contrast, NH₄⁺-N decreased under both salinity treatments relative to CK (p < 0.05), with comparable levels in LS and MS.

As shown in Figure 4, alfalfa growth responses differed among treatments at harvest. Plant height peaked under low salinity and was significantly higher than under moderate salinity (p < 0.05), whereas differences between the control and the two salinity treatments were not significant. In contrast, both fresh biomass and dry matter accumulation were significantly reduced under low and moderate salinity compared with the control (p < 0.05), with no significant difference between the two salinity levels. Together, these results indicate that mild salinity was associated with a height increase without a corresponding increase in biomass, while overall biomass production was sensitive to salinity stress.

As shown in Figure 7, nitrogen cycling–related functional profiles exhibited treatment-associated trend-level variation based on row-wise z-score–normalized relative abundances. Overall, several functional groups displayed heterogeneous enrichment/depletion patterns across replicates, with the most apparent variability observed in nitrate reduction–related functions and auxiliary nitrogen metabolism, whereas some categories (e.g., organic degradation/synthesis and anammox) showed comparatively stable or low signals across treatments. To formally evaluate treatment effects, Kruskal–Wallis tests followed by Dunn’s post hoc tests with FDR correction were performed on the underlying (non–z-score) relative abundance values. After FDR correction, no nitrogen cycling functional category differed significantly among treatments (FDR-adjusted p > 0.05). Therefore, Figure 7 is presented to visualize exploratory, relative patterning among samples rather than confirmatory, statistically supported directional changes.

Based on the KEGG level 3 pathway heatmap (Figure 8) generated from row-wise z-score–normalized relative abundances, microbial functional profiles showed treatment-associated trend-level patterning along the salinity gradient. Overall, pathways related to core carbon/energy metabolism tended to show relatively higher z-scores in CK and lower z-scores under salinity, particularly in MS, whereas LS often displayed intermediate profiles. In contrast, pathways associated with environmental sensing/regulation and transport functions showed heterogeneous responses across replicates, with some categories exhibiting relatively higher z-scores in LS than in CK and MS. To evaluate treatment effects statistically, pathway-level differences were tested using Kruskal–Wallis tests followed by Dunn’s post hoc tests with Benjamini–Hochberg FDR correction on the underlying (non–z-score) relative abundance values. After FDR correction, most pathways were not significant (FDR-adjusted p > 0.05), indicating that Figure 8 is presented primarily to visualize coordinated, exploratory trends in functional potential rather than confirmatory between-treatment differences.

The Mantel tests indicated that variation in genus-level microbial community composition was associated with variation in soil physicochemical properties across samples (Mantel r = 0.60, p = 0.005; Figure 9). In the Pearson correlation matrix, plant growth traits (plant height, fresh weight, and dry weight) were positively correlated with total nitrogen (TN) and NO₃^--N, but negatively correlated with NH₄⁺-N, whereas EC and pH were negatively correlated with biomass traits (Figure 9).

When Mantel associations were examined by treatment group, LS and MS displayed more detected links (p < 0.05) than CK, with pH, EC, and TN showing relatively stronger associations with community composition (Figure 9). Overall, these results suggest that shifts in soil salinity-related properties and nitrogen status are accompanied by coordinated changes in microbial community structure, while plant growth traits covary with soil nitrogen forms. Given the low replication, these correlation-based results are interpreted as exploratory associations rather than evidence of causality.

Salinity markedly altered soil physicochemical properties and soil inorganic nitrogen status, consistent with characteristic features of saline soils. Notably, although MS received a higher salt addition than LS, EC measured at harvest (30 Sep 2024) was slightly higher in LS. This pattern may reflect not only the initial salt dose but also in-season salt transport and redistribution driven by irrigation.

Under regular irrigation, salts can migrate with soil water flow, and leaching may move soluble ions downward, causing the salt content in the sampled layer at harvest to deviate from a linear dose–response relationship (Wang et al., 2024; He et al., 2025). In addition, because the salt source included SO₄²^-, sulfate may react with Ca²⁺ in soil solution and promote precipitation of CaSO₄ phases (e.g., gypsum), thereby reducing the dissolved-ion pool and EC, potentially more strongly under higher sulfate inputs (Van Driessche et al., 2019; Reiss et al., 2021; Ziegenheim et al., 2020). These mechanistic explanations remain speculative given that EC was assessed only at harvest; future work with time-series EC/major-ion measurements (e.g., Cl^-, SO₄²^-, Na⁺, Ca²⁺, Mg²⁺) and leachate monitoring would be required to test them.

Soil pH increased significantly from CK to LS and further to MS, indicating progressive alkalinization under salinity. This increase may be related to Na⁺-associated alkalinization processes and exchange reactions that promote OH^- accumulation. Although legumes can acidify the root-zone via organic acid exudation to facilitate nutrient acquisition (He et al., 2020), this buffering capacity may be weakened under salinity stress, which could contribute to the observed pH increase (Su et al., 2022).

Salinity also shifted soil nitrogen status. Total N increased from CK to MS, which could reflect reduced plant N uptake or altered microbial turnover under saline conditions. NO₃^--N was significantly elevated in MS, whereas CK and LS did not differ, suggesting that nitrate accumulation became more pronounced only under moderate salinity in this pot system. In contrast, NH₄⁺-N was lower under salinity treatments than in CK, consistent with changes in ammonium retention and/or transformation under more alkaline conditions. These shifts in nitrate-to-ammonium balance likely reflect a combination of plant uptake and microbial processes and could also be influenced by NH₄⁺ volatilization under higher pH (Min et al., 2014) and by salinity-related changes in leaching dynamics (Amini et al., 2016). In addition, salinity-induced inhibition of alfalfa nodulation has been reported (Aslani Borj et al., 2022) and may contribute to altered N inputs and nitrogen form distributions, although nodulation was not quantified in the present experiment.

Alfalfa growth traits responded differentially across the salinity gradient, with shoot elongation and biomass accumulation showing distinct sensitivities. Plant height reached its highest mean value under low salinity (LS) and was significantly higher than under moderate salinity (MS), whereas the difference between LS and CK was not statistically significant. This pattern is consistent with the possibility that mild salinity coincided with maintained shoot elongation capacity in this experiment. Potential explanations include osmotic adjustment and stress signaling processes reported in other systems (e.g., proline-associated responses) (Kong et al., 2020), although these mechanisms were not directly measured here. Root-derived organic acids may also influence the root-zone chemical environment and could alleviate early constraints on growth (Fan et al., 2023). Under MS, plant height was numerically lower than CK and significantly lower than LS, consistent with stronger ionic stress and Na⁺-related disruption of K⁺ uptake constraining growth processes (Zhang et al., 2010), although direct ion measurements in plant tissues would be required to confirm this interpretation.

In contrast to plant height, biomass accumulation was more consistently reduced under salinity. Both LS and MS had significantly lower fresh and dry biomass than CK, while differences between LS and MS were not significant, indicating that biomass production was sensitive to salinity even when height differences were modest. The decline in fresh weight likely reflects osmotic constraints on plant water status, whereas reduced dry matter suggests limitations on carbon assimilation and growth efficiency. These patterns are consistent with reported salinity-associated impacts on photosynthesis (e.g., chloroplast damage and reduced Rubisco activity) (Noori et al., 2018) and ionic imbalance effects on protein synthesis and enzyme function (Li et al., 2010). Alfalfa can adjust biomass allocation (e.g., root:shoot ratio) under salinity (Liu et al., 2011), which may buffer some traits; however, the overall biomass reduction suggests that any compensation was insufficient to offset stress effects under the conditions tested.

In this pot experiment, salinity was associated with shifts in root-zone microbial community composition (β-diversity), whereas α-diversity metrics showed limited treatment effects. Shannon and Simpson indices at the genus and species levels did not differ among treatments, and the clearest α-diversity signal was observed for the KEGG-entry Simpson index. This pattern is consistent with the possibility that salinity influences functional evenness more detectably than taxonomic evenness at the tested sequencing depth. Such observations align with previous reports that salinity can reshape microbial composition and activity (Sunita et al., 2020), while functional redundancy may buffer diversity responses at broader functional levels (Debray et al., 2022; Wahab et al., 2023). Recent syntheses, including a meta-analysis across rhizosphere studies, similarly highlight salinity as a major driver of root-associated microbiome assembly across diverse plant systems (Abdelfadil et al., 2024);.

At the taxonomic level, several dominant genera exhibited treatment-associated abundance tendencies (e.g., lower relative abundance of Nocardioides and higher relative abundance of Bradyrhizobium and Sediminibacterium under salinity). However, none of the top 10 genera differed significantly among treatments after FDR correction; therefore, genus-level differences should be interpreted cautiously as trend-level patterns. Even so, the observed directions are compatible with the concept that salinity can act as an environmental filter through osmotic stress and ion toxicity, resulting in taxon-specific responses (Chen et al., 2024; Reginato et al., 2021). Moreover, mixed-salt inputs (NaCl and Na₂SO₄) may exert ion-specific effects on soil processes and biota (Irakoze et al., 2021; Reginato et al., 2021), potentially contributing to heterogeneous genus-level tendencies. Reports of salt-tolerant rhizobia (including Bradyrhizobium) supporting legume performance under salinity provide a plausible context for the observed tendency of Bradyrhizobium in this dataset (Dong et al., 2017), although symbiotic performance was not directly assessed here.

For nitrogen-cycling categories and KEGG metabolic pathways, heatmaps showed treatment-associated patterning in row-wise z-score profiles, with LS exhibiting distinct profiles for some functions relative to CK and MS. Importantly, formal tests on the underlying (non–z-score) relative abundance values did not yield significant between-treatment differences after FDR correction, indicating limited statistical support for pathway-specific claims. Accordingly, these functional results are best interpreted as coordinated, subtle trends in metagenome-inferred functional potential that may motivate targeted validation (e.g., process measurements, enzyme assays, or transcript-based analyses), rather than confirmatory evidence for shifts in specific nitrogen transformation pathways. Within these constraints, the observed patterns remain broadly consistent with prior work suggesting salinity effects on nitrogen transformation guilds (Ouni et al., 2014; Zhu et al., 2023) and stress-response strategies (e.g., transport and regulatory systems) in salt-affected environments (Waheed et al., 2022; Shilev, 2020). Overall, the present results are compatible with the view that environmental filtering and plant–microbe interactions may contribute to community reorganization under salinity (Quiroga et al., 2017; Zhang et al., 2019; Dastogeer et al., 2020; Yu et al., 2021), while highlighting that many apparent differences warrant validation in larger-scale studies given the low replication and multiple-testing correction (Nuccio et al., 2013).

The observed covariation between salinity-related soil properties (e.g., EC) and plant biomass is consistent with widely reported salinity effects on plant water relations and ion balance. Mild salinity can sometimes coincide with short-term physiological adjustment (e.g., osmolyte accumulation and ion compartmentalization), whereas stronger or prolonged salinity is frequently linked to disrupted Na⁺/K⁺ homeostasis and growth inhibition (Kumar et al., 2021; Bano et al., 2021). In our dataset, plant growth traits also covaried with soil nitrogen forms, suggesting that nitrogen availability/status may be relevant to alfalfa performance under the tested conditions. Recent reviews also emphasize that plant–microbe interactions in the rhizosphere can modulate plant ion balance and stress responses under salinity, supporting the use of integrative soil–plant–microbiome frameworks for hypothesis generation (Ren et al., 2026).

Previous studies have reported that plant-associated symbioses (e.g., AMF) and beneficial microbial assemblages can be linked to improved nutrient acquisition and stress tolerance, and that root exudation may shape microbial community assembly under abiotic stress (Diagne et al., 2020; Hajihashemi et al., 2020; Liu et al., 2020). In contrast, weaker or inconsistent relationships involving NH₄⁺–N could reflect multiple non-exclusive processes, including salinity impacts on nitrification, plant N uptake preference, and reduced NH₄⁺ availability under higher pH due to volatilization losses (Mansour and Hassan, 2022; Kumar et al., 2021). Within this conceptual framework, enrichment of salt-tolerant taxa with stress-adaptive traits may accompany increases in EC, while mycorrhizal symbiosis and root-zone signaling processes have been proposed to support osmotic regulation and water balance (Sharma et al., 2019; Ramadhani and Widawati, 2020). Collectively, these literature-supported mechanisms provide hypotheses that are broadly consistent with the correlation patterns observed here and offer a conceptual basis for interpreting coordinated soil–microbe–plant responses along a salinity gradient (Hu et al., 2018; He et al., 2021a, He et al., 2021b, He et al., 2022).

Importantly, the present study provides associational evidence linking plant traits with soil chemistry and microbial functional profiles. Demonstrating causal mediation would require additional experiments (e.g., microbial inoculation, sterilized soil reconstitution, or targeted manipulation of key functions), ideally with larger replication and field-scale validation.

Several limitations should be considered. First, biological replication was low (n = 3 per treatment), which limits statistical power-particularly for multivariate analyses and functional comparisons after FDR correction-and increases uncertainty around subtle effects. Second, this was a pot experiment with restricted rooting volume and simplified water–salt dynamics; therefore, extrapolation to field-scale saline soils should be made cautiously. Third, shotgun metagenomics infers functional potential rather than directly measuring nitrogen transformation rates; additional process-based measurements (e.g., N transformation rates, enzyme activities, or metatranscriptomics) are needed to verify whether inferred functional trends translate into realized biogeochemical changes. Future studies with larger replication and field validation across saline gradients will be important to confirm the observed soil–microbiome–plant associations and evaluate their agronomic relevance.