Single‐cell Raman and functional gene analyses reveal microbial P solubilization in agriculture waste‐modified soils

Abstract Application of agricultural waste such as rapeseed meal (RM) is regarded as a sustainable way to improve soil phosphorus (P) availability by direct nutrient supply and stimulation of native phosphate‐solubilizing microorganisms (PSMs) in soils. However, exploration of the in situ microbial P solubilizing function in soils remains a challenge. Here, by applying both phenotype‐based single‐cell Raman with D2O labeling (Raman‐D2O) and genotype‐based high‐throughput chips targeting carbon, nitrogen and P (CNP) functional genes, the effect of RM application on microbial P solubilization in three typical farmland soils was investigated. The abundances of PSMs increased in two alkaline soils after RM application identified by single‐cell Raman D2O. RM application reduced the diversity of bacterial communities and increased the abundance of a few bacteria with reported P solubilization function. Genotypic analysis indicated that RM addition generally increased the relative abundance of CNP functional genes. A correlation analysis of the abundance of active PSMs with the abundance of soil microbes or functional genes was carried out to decipher the linkage between the phenotype and genotype of PSMs. Myxococcota and C degradation genes were found to potentially contribute to the enhanced microbial P release following RM application. This work provides important new insights into the in situ function of soil PSMs. It will lead to better harnessing of agricultural waste to mobilize soil legacy P and mitigate the P crisis.


INTRODUCTION
Phosphorus (P) is an essential element for agriculture production but has the most limited bioavailability in soils compared to other plant nutrients 1 .Large amounts of P fertilizers have been applied to soils, but only a small fraction (around 10%) of P is bioavailable for crop growth, with most of the P being fixed in soils in either insoluble organic or inorganic forms [2][3][4] .This leads to a paradoxical situation wherein the soils are rich in fixed-P but the plants are deficient in P. Mineral P fertilizer derived from phosphate rocks is a nonrenewable resource that has been predicted to be depleted by 2100 5 .Mobilization of the soil fixed-P for agricultural production provides a sustainable way to reduce P fertilization and mitigate the P crisis.Soil microorganisms play a crucial role in mobilizing recalcitrant soil P. Naturally occurring phosphate-solubilizing microorganisms (PSMs) in soils can secrete phosphatases to mineralize organic P or produce organic acids to solubilize inorganic fixed-P [6][7][8] .Moreover, amendments of soils with organic nutrients, such as carbon (cellulose, lignin, and glucose), organic fertilizers, and crop straw, can enhance P turnover by regulating both soil chemical (C, N, and P) and microbial properties (PSM population and enzyme activity) 9 .These organic fertilizers such as rice straw and livestock manure are presently encouraged to apply to soils for sustainable agricultural production 10 .Rapeseed meal (RM), a by-product of rapeseed oil extract, is an environmentally friendly agriculture waste.It has richer nutrients compared to manure fertilizer and has great potential for use toward improving soil quality and driving microbial mobilization of soil-fixed P 11 .However, the effects of RM application on soil P pool, microbial Psolubilization activity, and the interaction among the functional genes are largely unknown.
Despite the important role of PSMs in mobilizing soil fixed-P, it remains a major challenge to study the in situ function of PSMs in native soils because the vast majority of soil bacteria are still uncultured 12 .Presently, neither the physiology studies of PSM isolates based on culture-dependent methods nor genomic databases on culture-independent metagenome or PCR can reliably predict the microbial function and activity in their native habitat.Single-cell Raman spectroscopy provides a novel culture-independent means of in situ phenotyping bacterial function and activity based on the intrinsic biomolecules of microbial cells 13,14 .When combined with stable isotope probing (SIP) such as 13 C, 15 N, and 2 deuterium (D), remarkable Raman shifts induced by SIP assimilation can act as biomarkers for diverse microbial functions, such as N 2 /CO 2 fixation, drought tolerance, and antibiotic resistance 13,[15][16][17][18][19] .Recently, Li et al. 18 developed single-cell Raman with D 2 O labeling (Raman D 2 O) as a new activity-based approach to identify active PSMs in native soils.It is based on the finding that PSMs are more active in utilizing fixed P for their metabolism by incorporating more D for de novo synthesis than non-PSMs.The generated C-D Raman band intensities displayed a linear relationship with microbial P solubilization ability, enabling further quantification of the P solubilization activity of individual cells.This new Raman D 2 O provides an important phenotypic approach to quantifying the abundance and in situ activity of PSMs in soils amended with RM agriculture wastes.
In addition to the phenotype characterization, it is also important to understand the underlying P solubilization mechanisms.P cycling is not independent but interacts with C and N cycling.The concentrations of soil organic C were found to regulate P cycling-related functional genes in agroecosystems because organic C mineralization is usually accompanied by hydrolysis of organic P 20 .In addition, N input significantly increased the abundance of PSMs and the microbial phosphatase activity 21 .The recently developed quantitative microbial element cycling (QMEC) smart chip can target 71 microbial CNP functional genes, including organic P mineralization genes (e.g., bpp, pccA, and phnK), inorganic P solubilization genes (e.g., pqqC, ppK, and pmoA), C degradation genes (e.g., amyA, xylA, and hzsB), and N nitrification genes (e.g., amoA and gdh) 22 .As these functional genes are related to fixed P release, quantification of the abundance of CNP functional genes is a useful way to understand how soil microorganisms respond to different fertilization strategies 23 .By using the QMEC chip, the abundances of CNP functional genes have been demonstrated to significantly increase in soils with a high starter P fertilization 24 .It is anticipated that coupling the phenotypic single-cell Raman approach with the genotypic functional gene detection will provide a comprehensive understanding of the impact of RM on microbial P solubilization in soils.
In this study, we established a microcosm experiment in which RM was added to three typical farmland soils.Both phenotypic single-cell Raman D 2 O and genotypic CNP gene chips were applied to study the microbial P cycling in these agriculture waste-amended soils.Together with determination of soil properties and soil microbiome sequencing, we aimed to (1) evaluate the effect of RM on soil properties, Psolubilizing activities of native soil microbes, and soil type dependence; (2) reveal the associated soil microbiome involved in P solubilization; and (3) decipher the CNP cycling genes and interaction between C, N, and P functional genes.This study provides important insights into the use of agriculture waste to promote the release of soil-fixed P by functional soil microbes, thereby leading to mitigation of the P crisis and a sustainable agriculture.

Effect of RM addition on soil and RM properties analysis
Soil samples of Dezhou (DZ) and Donghu (DH) soils were detected to be alkaline (pH > 7.5, Figure 1A,D), while Qiyang (QY) soils were acidic (pH < 6.5, Figure 1A,D).The concentrations of total P of soils and the properties of RM are shown in Tables S1 and S2.The concentrations of total P in DH soils were significantly higher than that in DZ and QY soils.A 60day microcosm experiment was designed to stabilize the soil microbial community after RM addition.The RM addition was found to exert no effect on soil pH in any of the treatments (p > 0.05, Figure 1A,D), while the concentrations of both Olsen P and dissolved organic carbon (DOC) were significantly increased in three RM-treated soils at both sampling times of 30 and 60 days (p < 0.001, Figure 1B,C,E,F).

Effect of RM addition on the phenotypic microbial P solubilization function
After adding D 2 O to the three soils from which soluble P was removed, single-cell Raman spectroscopy was used to study the PSMs that can actively mobilize fixed P in soils (Figure 2A).A total of over 100 single cells extracted from each soil were measured.Figure 2B shows the typical spectra of individual PSMs and non-PSMs with distinct C-D band intensities due to their different abilities of metabolizing D in only fixed P-containing soils.Figure 2C shows the C-D ratios of all the measured individual bacteria from the three soils at the second sampling time.Vast distributions of C-D ratios were observed, indicating the high activity heterogeneity of soil microbes in solubilizing fixed P. Additionally, RM application significantly increased the metabolic activity of PSMs in DH soils (p < 0.0001), but not in DZ and QY soils (p > 0.05).The abundance of PSMs was calculated by dividing the number of PSMs by the total number of Ramandetected bacteria (Figure 2D).We found that RM application significantly increased the abundance of PSMs in DH (p < 0.05) and DZ soils (p < 0.001) but not in QY soils (p > 0.05).These results indicated the soil-type-dependent effect of RM application on the phenotypic microbial P solubilization.The nonsignificant effect of RM on the QY soils may be related to its low pH, which could inhibit bacterial metabolism activity.It is noteworthy that single-cell Raman D 2 O enabled the direct examination of individual PSMs and in situ Psolubilization activity in native soils, which is not achievable using traditional bulk phenotypic methods such as soil enzyme activity and soil respiration rate.

Effect of RM addition on bacterial communities
To further understand the changes of soil microbial communities impacted by RM, a total of 1,263,169 high-quality sequences were obtained after merging and quality filtering from 36 samples, with each sample ranging from 27,326 to 45,496.Actinobacteriota, Chloroflexi, Gemmatimonadota, Bacteroidota, Proteobacteria, Acidobacteriota, and Firmicutes were the seven dominant phyla in all soil samples (Figure 3A).A higher relative abundance of Acidobacteriota was observed in RM treatment samples compared to the control (p < 0.05).No significant effect of RM application on bacterial alpha diversity (Chao 1 index) was observed in soil samples at the first sampling time (Figure 3B, p > 0.05), while RM application significantly decreased the alpha diversity at the second sampling time in all three soils (Figure 3C, p < 0.001), indicating that the rich nutrient input of RM decreased microbial diversity by potentially increasing the interspecific competition and enriching some species.Principal co-ordinates analysis (PCoA) with the Bray-Curtis distance revealed no significant difference in the bacterial communities with RM application (Figure S1, p > 0.05).Bacterial co-occurrence network was further applied to analyze bacterial community associations (Figure S2).The average clustering coefficient of the network was found to increase with RM application by 0.119.More edges and labels were observed in RM treatments compared to the control (Figure S2).These results indicated that RM application enhanced the interaction of bacterial community (relative abundance > 0.1%).The sensitive taxa after RM application in three soils were analyzed by linear discriminant analysis effect size (LEfSe) (p < 0.05, LDA > 3.0, Figures S3-S5).The results revealed multiple bacteria as the biomarkers sensitive to RM application.For example, genus_Acitinomadura and class_Myxococcia were sensitive to RM addition in DZ, DH, and QY soils.The results indicated that RM application could enrich some bacteria in soils.

Effect of RM addition on functional gene profiles
A total of 61 CNP cycling-related functional genes were detected in all soil samples (Figure 4A).These CNP genes were divided into seven groups based on the main element cycling involved, that is, C cycling genes (C degradation, C fixation, and methane metabolism), N cycling genes (nitrification and denitrification), and P cycling genes (organic P mineralization and inorganic P biosynthesis).Both soil types and RM application were found to affect the relative abundance of CNP functional genes (Figure 4B-G), and RM application also changed the composition of CNP genes in all soil samples (p < 0.05, Figure S6).In general, the abundances of CNP functional genes in RM-amended soils were significantly higher than those in the control (Figure 4B-G).More specifically, organic P mineralization and inorganic P solubilization genes were significantly increased following RM addition in DH and DZ2 soils (p < 0.01, Figure 4B,C,E), but not in QY2 soils at the second sampling time (p > 0.05, Figure 4G).This result is consistent with the increased abundance of phenotypic PSMs in DZ2 and DH2 soils observed by Raman D 2 O (Figure 2D).C cycling-related genes responsible for starch, hemicellulose, and cellulose degradation were detected in all samples.With the RM addition, the relative abundance of C degradation genes increased in almost all soils (p < 0.01, Figure 4B, D-F), except DH and QY soils at the second sampling time, suggesting that RM addition can potentially enhance carbon-degrading processes.The abundance of C fixation genes was much higher than C degradation genes, and significantly increased with RM addition in most soil samples (p < 0.01, Figure 4B,D-F).The abundance of nitrogen cycling genes also increased in most soil samples with RM addition (p < 0.05, Figure 4B,C,E,F).Furthermore, redundancy analysis (RDA) of the correlation between CNP functional gene profiles and environmental factors indicated that pH, DOC, and Olsen P were positively correlated with CNP functional gene profiles in all soil samples (Figure S7).Procrustes analysis and Mantel test revealed that bacterial communities were significantly correlated with CNP functional genes in the DZ and DH soils (p < 0.001, Figure 5A,B), but not in QY soils (p > 0.05, Figure 5C).

Association of microbial P solubilization ability with soil microbiome and functional genes
The associations between phenotypes (abundance of active PSMs) and genotypes (soil microbiome and functional genes) were further analyzed.Regarding the correlations between the abundance of soil microbiome at the phylum level and the abundance of active PSMs, Myxococcota, Bacteroidota, Dependentiae, and Bdellovibrionota (correlation index > 0.5), which were previously reported to be capable of inorganic P solubilization, were positively correlated with the proportion of active PSMs (Figure 6A).Among them, Myxococcota was positively and significantly associated with the proportion of PSMs (p < 0.05, R 2 = 0.467, Figure 6B).In addition, the functional genes of rbcL, nosZ1, and gam (correlation index > 0.5, Figure 6C), which encode starch, pectin, and lignin hydrolysis, respectively, were positively correlated with the abundance of phenotypic PSMs.

Effect of RM amendments on soil properties and P-solubilizing activities of soil microbiome
This study demonstrated that RM application not only improved soil quality by increasing soil DOC and Olsen P in all three soil types but also enhanced soil type-dependent microbial P solubilization activity.RM is a rich nutrient source that contains high contents of P, free amino nitrogen, and carbon, explaining the improvement of soil C and P contents following RM addition [25][26][27] .In addition, single-cell Raman D 2 O was applied to directly identify phenotypically active PSMs in native soils and quantify their abundance in a culture-independent way.A clear soil type-dependent effect of RM addition on microbial P solubilization was observed, that is, RM application significantly increased the abundance of active PSMs in two alkaline DZ and DH soils but exerted no effect on acid QY soils.The accelerated microbial P solubilization could be related to the carbon input from RM.Previous work has found that cellulose, hemicelluloses, and lignin significantly increased the abundance of PSMs and their activity of phosphatase 28,29 .For example, polysaccharide breakdown and sugar metabolism are important for stimulating phosphate solubilization 30 .In addition, organic fertilizer was found to increase the abundance of transcriptionally active phoD-harboring bacteria 31,32 .In our work, C degradation genes were found to increase with RM treatments, highlighting that the potential involvement of PSMs in carbon degradation may affect the metabolic traits of PSMs.In addition, considering that C mineralization usually couples with organic P hydrolysis, the addition of organic matter enhanced microbial capacity for C mineralization and P hydrolysis to maintain a balanced microbial C:P ratio in nutrient rich soils 33 .RM addition has been shown here to increase DOC in soils, thus stimulating the active PSM population.While this effects is soil type dependent.We found that RM addition increased the abundance of PSMs in alkaline DH and DZ soils but had no effect in acidic QY soils.Previous work revealed that the abundance of soil PSMs decreased with the decrease of soil pH 34,35 .The reason may be that the low pH restrained the promoting effect of nutrient input on bacterial activity 36,37 .It is noteworthy that previous studies focused on microbial P solubilization function were mainly based on bulk enzyme analysis.However, this method averages out microbial P solubilization activity but could not reveal the active PSM subpopulation.In this study, Raman D 2 O achieved direct identification of active PSMs in native soils and their activity heterogeneity at the high single-cell resolution.The PSMs with high activity of releasing fixed P and utilizing complex organic matter have great potential as bacterial fertilizers.Future studies combining single-cell sorting and cultivation of key active PSMs will help to explore novel functional bacterial resources to alleviate the P crisis.

Soil microbiome involved in P solubilization in three RM-amended soils
Bacterial diversity is a fundamental factor in the soil ecosystem and plays an important role in nutrient cycling.Our study found that RM application significantly decreased the diversity of soil bacteria.The increased DOC and Olsen P following RM application change the soil substrate pool, which may stimulate the proliferation of RM utilizers but inhibit non-RM utilizers, especially for the microorganisms that tend to live in oligotrophic habitats, thereby reducing the diversity of soil microorganisms 38 .Moreover, PCoA analysis revealed no significant difference in the bacterial community structure, suggesting that only the lowabundance bacteria were eliminated through RM application without affecting the structure of high-abundance bacteria.Similar effects of nutrient input on soil bacterial communities were also reported previously 39 .Co-occurrence network analysis visualizing the scenarios of bacterial interactions showed that more nodes, edges, and average clustering coefficients were observed in RM-treated soils.This result indicated that the application of RM enhances the interaction among soil microbiome.The more negative interactions of soil microorganisms with RM addition may enhanced bacterial interactions to promote their resistance to the change of soil substrate 40 .
The LEfSe analysis revealed some RM-sensitive bacteria in three soils, including Bacillus, Actinomadura, and Myxococcia, which have been reported as keystone bacteria in agriculture ecosystems linked to the P solubilization function [41][42][43] .A further correlation analysis of the Raman-based phenotypes with the taxonomy-based genotypes revealed that the abundance of Myxococcota was significantly positively correlated with the proportion of active PSMs.Meanwhile, Myxococcota was also a sensitive biomarker for RM addition.This phylum is known to play an essential role in carbon cycling that could produce organic acid to promote P solubilization 44 .These results indicated that Myxococcota may be the potential active PSMs contributing to the solubilization of inorganic fixed P in soils.

CNP cycling genes and the interaction between C, N, and P functional genes
To further explore the genotypic mechanisms of the observed microbial P solubilization in RM-amended soils, a total of 61 CNP functional genes were detected and quantified by QMEC gene chips.Our results revealed that RM application significantly altered the structure of microbial functional genes.The abundance of CNP genes in most soils increased with RM addition, especially those involved in C degradation, C fixation, nitrification, denitrification, organic P mineralization, and inorganic P solubilization.Moreover, the abundance of P cycling-related genes had significant positive linear correlations with the abundance of C and N cycling-related genes (Figure S8).These results indicate that microbial P releasing processes are not independent but interact with C and N cycling 45 .The possible reason is that some functional bacteria driving C and N cycling, such as C degradation and N nitrification, can also produce organic acid, ATP, and protons to enhance P cycling 24 .A further correlation analysis indicated that C degradation genes of rbcL, nosZ, and gam were positively correlated with the abundance of active PSMs, indicating that the organic acid secretion during C degradation may mainly account for the microbial solubilization of fixed P. This was consistent with the increase of active PSM abundance in soils containing a higher DOC content following RM amendments.As such, the potential CNP functional genes that contribute to the enhanced microbial P mobilization upon RM addition were deciphered.
In conclusion, this work applied both single-cell Raman D 2 O and CNP functional gene chips to reveal the phenotypes and genotypes of microbial P solubilization function in three types of farmland soils amended with agricultural waste of RM.RM addition increased both the soil C and P contents and the abundance of phenotypically active PSMs in native DZ and DH soils.Genotypic results indicated that RM application generally increased the abundance of CNP functional genes and enriched the abundance of P solubilization related bacteria.Further correlation analysis revealed that Myxococcota-and C degradation-related genes may mainly contribute to the enhanced microbial P solubilization function.This study clearly demonstrates that RM is an agricultural friendly product that not only improves soil quality but also, more importantly, improves microbial P solubilization activity.The application of two cultureindependent methods established a potential link between the phenotype and genotype of soil-native PSMs, which helped to explain the mechanisms about how RM addition increased the microbial P solubilization function.These findings will advance our understanding and harnessing of agricultural by-product fertilization for the sustainable development of agriculture.

Soil sample collection and incubation
Soil samples were collected from the farmlands located in DH: paddy soil, DZ: fluvo-aquic soil, and QY: red soil.The surface layer (0-15 cm) of soils was taken for our study.The collected soils were air-dried in the dark and the allogenic matter was removed by a 2 mm sieve.The RM was collected from a rap oil processing plant in Jiaxing, Zhejiang province.The RM was air-dried and ground to homogenize the particle size before application into soils.The abovementioned three soils were amended without (control) or with 1% RM (10 g of RM per kg soil).Each treatment was performed in triplicate.The incubation experiments were performed with 60 g of soil in 100 ml glass beakers.The soil moisture was adjusted to 75% of field capacity.The soils were incubated in an illumination incubator at 25°C and sampled after 30 and 60 days of incubation.A total of 18 soil samples were obtained, and each sample was divided into two parts.One part was stored at −20°C for DNA extraction and the other part was used for determination of soil properties and Raman detection.

Soil and RM property analyses
pH of soil and RM was measured after shaking soils in water at a soil to water ratio of 1:2.5 (w/v).Total P concentrations of soil and RM were determined by HClO 4 -H 2 SO 4 digestion, followed by molybdenum-blue colorimetric measurement 46 .Olsen P was extracted with a NaHCO 3 solution and detected using the molybdenum-blue colorimetric method 47 .Concentrations of dissolved organic carbon (DOC) were determined using a TOC analyzer (TOC-LCPH).

Single-cell Raman spectroscopy to probe active PSMs in soils
To obtain available P-free soil, the soil samples were washed three times with 0.5 mol/l NaHCO 3 solution and ultrapure water, respectively, until there was no obvious chromogenic reaction detected by the molybdenum-blue colorimetric method.Briefly, 1 g of soil was added to 30 ml of NaHCO 3 solution, incubated at 25°C and 180 rpm for 8 h, and then centrifuged to remove the supernatant.The same procedure was applied to the ultrapure water 18 .The obtained soils were then air-dried for 24 h.An aliquot of 200 μl of D 2 O was added to 1 g of soils to simulate soil moisture in nature.The soil samples were incubated at 37°C for 24 h before extraction of soil bacteria for Raman analysis.
Soil bacteria were extracted using a modified Nycodenz density gradient separation protocol 48 .Briefly, 1 g of soil was homogenized in 3 ml of phosphate-buffered saline (PBS) with 20 μl of Tween 20.The soil slurry was vortexed for 20 min to detach soil-associated bacteria.The vortexed slurry was then generally added to 3 ml of Nycodenz solution (Aladdin, 1.42 g/ml).To separate bacteria from soil particles, the slurry was centrifuged at 14,000 g for 90 min at 4°C.The middle layer containing bacteria was carefully collected into a clean 1.5 ml tube and washed with water to remove residual PBS.Two microliters of the obtained bacterial solution was loaded on an Al foil for single-cell Raman measurements.A LabRAM Aramis confocal Raman microscope (HORIBA Jobin-Yvon) was used for PSM identification.It is equipped with a 532 nm Nd:YAG laser (Laser Quantum), 300 g/mm grating, and a 100× objective (Olympus, NA = 0.09).For each bacterium, the spectrum in the range of 500-3200 cm −1 was obtained at an acquisition time of 9 s.All spectra were analyzed using LabSpec5 software (HORIBA Jobin-Yvon) for baseline correction and peak intensity quantification.C-D ratios of CD/ (CD + CH) were calculated to quantify the microbial activity for P solubilization.The C-H and C-D Raman peaks used for this calculation were at 2800-3100 and 2040-2300 cm −1 , respectively.
Soil DNA extraction, amplification, highthroughput sequencing, and analysis Soil DNA was extracted using the DNeasy PowerSoil Kit (QIAGEN).The quality and concentration of DNA were monitored using a Qubit 4 fluorometer and agarose gel electrophoresis.The bacterial 16S rRNA gene was amplified using primers of 515F-806R according to the previously reported amplification conditions 49 .The obtained gene products were sequenced on an Illumina MiSeq PE 300 platform (Meiji Biological Medicine Co).Quantitative Insights Into Microbial Ecology (QIIME, version 1.9.1) was chosen to analyze the high-throughput sequencing data 50 .Low-quality reads and primer sequences were removed to obtain clean sequences.Only the sequences with over 97% of similarity were assigned to operational taxonomic units (OTUs) by using UCLUST 51 , and taxonomic assignment to OTUs was based on the Greengenes database.

High-throughput qPCR for quantifying CNP cycling functional genes
To reveal the effect of RM application on CNP cycling, a total of 72 primer sets targeting 71 functional genes involved in CNP cycling and 16S rRNA genes were measured using the QMEC chip by high-throughput qPCR (HT-qPCR) 22 .HT-qPCR was performed on the Wafergen SmartChip Real-time PCR system (Wafergen Inc.) and three technical replicates were set for each sample.The information on primers is shown in Table S3.SmartChip qPCR software (V 2.7.0.1) was used to analyze the HT-qPCR data.The cutoff value of threshold cycle (C t ) was set to 31 and the data with amplification efficiencies higher than 0.9-1.1 were discarded.The relative abundance of functional genes obtained by normalizing functional gene copy numbers with the 16S rRNA gene copy numbers was used in our study.

Statistical analysis
Adonis analysis was used to analyze the similarities and differences of bacterial composition and presented by nonmetric multidimensional scaling plots (NMDS) using the Bray-Curtis dissimilarity matrix.Significant differences in soil chemical properties, the relative abundance of CNP functional genes, and the abundance of active PSMs were determined by one-way analysis of variance.Gene cooccurrence networks were constructed based on the Spearman's correlation matrix and visualized with Gephi V.0.9.2.Soil bacterial biomarkers (from the phylum level to the genus level) that were sensitive to RM addition were determined by LEfSe, and the threshold of the logarithmic LDA score was set as 3.0.Spearman's rank correlations between the proportion of PSMs and the abundance of bacteria at the phylum level and functional genes were calculated using the R function "cor.test.".Origin 2021, R and DataGraph were used to plot the graphs.

Figure 1 .
Figure1.Changes in soil properties following rapeseed meal (RM) amendment.Effects of RM addition on soil pH (A, D), Olsen P concentration (B, E), and dissolved organic carbon (DOC) (C, F).Data from three replicates (n = 3) are represented as means ± SD.One-way analysis of variance was used to test the significant difference between samples.DH, DZ and QY represented the soils sampled from Donghu, Dezhou and Qiyang, respectively.The numbers 1 and 2 represent the first (30 days) and the second (60 days) sampling time, respectively.***p < 0.001; ns, no significance.

Figure 2 .Figure 3 .
Figure 2. The effect of RM addition on phenotypic microbial P solubilization function.(A) Workflow for in situ phosphate-solubilizing microorganism (PSM) identification via single-cell Raman D 2 O. (B) Typical single-cell Raman spectra of PSMs and non-PSMs.(C) Distribution of C-D ratios measured from over 100 randomly selected single cells in DH, DZ, and QY soils amended with and without RM addition.Each point shows a measurement of a single cell.The red line at 10% shows the threshold for PSM identification.It was calculated as the mean + 3 × SD of C−D ratios from randomly selected bacteria incubated without D 2 O. (D) The effects of RM addition on the abundance of PSMs in DH, DZ, and QY soils identified by single-cell Raman D 2 O. *p < 0.05, ***p < 0.001, ****p < 0.0001; ns, no significance.

4 .Figure 5 .
Figure 5.The correlation between soil bacterial communities and functional gene profiles.Procrustes analysis and Mantel test reveal the correlation between CNP functional genes and bacterial communities on the basis of Bray-Curtis dissimilarity metrics in DH (A), DZ (B), and QY (C) soils, respectively.

Figure 6 .
Figure 6.The correlations between bacterial communities, functional gene profiles, and the abundance of phosphate-solubilizing microorganisms (PSMs).(A) Correlations between the relative abundances of bacteria at the phylum level and the proportions of active PSMs (*p < 0.05).(B) Regression relationships between the abundances of Myxococcota and the proportions of active PSMs.(C) Correlations between the relative abundances of functional genes and the proportions of PSMs.