The aim of the present study was to investigate the effect of a synthetic microbial consortium (SMC) containing five functionally different bacterial strains (Rahnella aquatilis, Rothia endophytica, Stenotrophomonas indicatrix, Burkholderia contaminans, Lelliotia amnigena) on the growth and development of three agricultural crops (wheat, buckwheat, and rapeseed) on two soil types (chernozem and gray forest soil) under field conditions. The experiment was conducted from June to September 2024 under extreme field conditions, with temperatures reaching 43.8 °C. This study evaluates SMC efficacy under severe abiotic stress, reflecting increasingly common climate extremes. Metagenomic data analysis showed that the introduced strains did not establish stable populations in the soil, possibly due to heat-induced bacterial mortality, though other factors including competition with indigenous microflora and lack of protective formulations may have also contributed. No statistically significant effects on plant morphometric parameters were observed. The extreme temperature and water stress conditions appear to have been the dominant limiting factors, overriding any potential benefits from microbial inoculation, as evidenced by the lack of response to mineral fertilizer application as well. Crop-specific effects were revealed: when cultivating rapeseed on chernozem, a significant increase in available phosphorus content was noted (from 278 ± 45 to 638 ± 92 mg/kg with SMC application, p < 0.001).

Plant growth-promoting bacteria (PGPB) are increasingly deployed as biofertilizers, yet the link between an inoculant’s genomic potential and its realized effect on the plant is rarely assessed within an integrative framework that jointly captures the rhizosphere microbiome, plant phenotype, and strain genome. Two Bacillus strains—B. halotolerans 1453 and B. pumilus 630—were applied to wheat and soybean in a factorial pot experiment (2 strains × 2 application methods × 3 frequencies + control, 3–4 replicates). Rhizosphere samples (n = 67 after filtering) were profiled by 16S rRNA sequencing with PICRUSt2 functional prediction and compositional validation (Aitchison PERMANOVA, ALDEx2, ANCOM-BC2). The PGPB gene repertoire was characterized by genome mining (481 marker genes, 14 categories). Wheat phenotype (six traits) and soybean height were analyzed with models appropriate for count data (Negative Binomial and binomial GLMs) for treatment-vs.-control comparisons, and with factorial ANOVA for decomposition into main effects and interactions. Crop identity was the dominant factor shaping both microbiome structure and function (PERMANOVA R² = 14.7% taxonomically and R² = 7.8% functionally, both p < 0.001), with biologically meaningful taxonomic differences between wheat and soybean; strain, application count and method had no significant effect on community composition (R² < 4% each), and co-occurrence networks showed no reliable differences between crops once read depth and sample size were controlled for. Despite this neutrality at the microbiome level, inoculation significantly increased wheat spike count (NB-GLM, all 12 treatments vs. control, padj 0.0002–0.031), ear weight, and stem count, with application count the strongest source of variability and a pronounced strain × application count. Strain 1453 outperformed 630 in spike count (+23.1%, p = 0.012) and ear weight (+20.4%, p = 0.023); we hypothesize that this may be related to its more complete DNRA pathway (narGHI + nirB-nirD) and biocontrol genes (bacE, srfAA). Strain 630 produced a less pronounced effect than strain 1453 but was subject to smaller fluctuations across replicates (CV ≈ 16–21% vs. ≈24–26% for 1453), which may reflect better resilience to environmental fluctuations, possibly due to its confirmed rsbV/rsbW stress-tolerance regulon. Rhizosphere microbiome composition differed clearly by crop (wheat vs. soybean) but showed no detectable response to strain, application method, or application count. Despite this lack of a microbiome signal, inoculation significantly increased wheat spike count and ear weight, with the magnitude and stability of this effect differing by strain. We hypothesize that this strain-dependent difference relates to underlying genomic differences—particularly in nitrogen metabolism (DNRA pathway) and stress-tolerance genes—though this link has not been tested directly and remains a hypothesis for future work.