Brassica seed meal fumigation restores beneficial bacterial communities by enriching taxa with high resistance and resilience

Brassica seed meals (BSMs) are widely used as biofumigants to control various soil-borne plant diseases. However, the mechanism of bacterial community reconstruction following fumigation with BSMs remains poorly understood. In the present study, to decipher the mechanism of bacterial community reconstruction in fumigated soil and to understand the effect of bacterial communities in fumigated soil on rhizosphere recruitment and subsequent disease control, we investigated the bulk soil and rhizosphere bacterial communities using field experiments in soils treated with various fumigants (Brassica campestris seed meal, Brassica juncea seed meal, and the chemical fumigant dazomet) in two greenhouses subsequently used to grow tomato and cantaloupe, respectively. This study revealed that bulk soil bacterial community composition changed significantly after fumigation extinction and recovery. Firmicutes and Proteobacteria, showing high resistance and resilience in the extinction and recovery processes, respectively, represented the key microorganisms for community reconstruction and rhizosphere recruitment. Moreover, nutrients supplied by BSMs, especially available phosphorus (AP), in fumigated soils determined the post-restoration changes in bacterial community composition. Additionally, BSMs showed greater potential than chemical fumigant dazomet in rebuilding beneficial bacterial communities and controlling potential soil pathogens by enriching gram-positive phyla Firmicutes and Actinobacteria and their respective affiliated genera Bacillus and Streptomyces. These results provide a fundamental ecological understanding of the response of soil-inhabiting microbes to fumigation and the reconstruction of soil beneficial bacterial communities after BSM fumigation. Thus, this study improves the understanding of the effects of biofumigants on soil-borne plant disease control in agriculture.

recruitment and subsequent disease control, we investigated the bulk soil and rhizosphere bacterial communities using field experiments in soils treated with various fumigants (Brassica campestris seed meal, Brassica juncea seed meal, and the chemical fumigant dazomet) in two greenhouses subsequently used to grow tomato and cantaloupe, respectively. Results This study revealed that bulk soil bacterial community composition changed significantly after fumigation extinction and recovery. Firmicutes and Proteobacteria, showing high resistance and resilience in the extinction and recovery processes, respectively, represented the key microorganisms for community reconstruction and rhizosphere recruitment. Moreover, nutrients supplied by BSMs, especially available phosphorus (AP), in fumigated soils determined the post-restoration changes in bacterial community composition. Additionally, BSMs showed greater potential than chemical fumigant dazomet in rebuilding beneficial bacterial communities and controlling potential soil pathogens by enriching gram-positive phyla Firmicutes and Actinobacteria and their respective affiliated genera Bacillus and Streptomyces. Conclusion These results provide a fundamental ecological understanding of the response of soilinhabiting microbes to fumigation and the reconstruction of soil beneficial bacterial communities after BSM fumigation. Thus, this study improves the understanding of the effects of biofumigants on soilborne plant disease control in agriculture.

Introduction
Soil microbial community and diversity play key roles in maintaining sustainable agroecosystems and crop production. However, agricultural intensification and unsustainable soil management practices have worsened microbial deterioration and soil-borne diseases . Crop diseases caused by soil-borne pathogens have been widely recognized as a serious problem in agricultural production systems due to the continuous monocropping and the abuse of chemical fertilizers (De Corato 2020; Ren et al. 2018;Shen et al. 2021). In agriculture management, fumigants such as dazomet (3,5-dimethyl-1,3,5-thiadiazinane-2-thione), Telone (1,3-Dichloropropene), methyl isothiocyanate (MITC), and chloropicrin (CP) are widely used as soil amendments before planting to control soil-borne diseases (Ajwa et al. 2010;Fu et al. 2012;Ibekwe et al. 2001;Zasada et al. 2010). However, chemical fumigants have been reported to have environmental legacy issues (Bell 2000), for example, fumigation with chloropicrin affects nitrogen cycling in soil and negatively impacts soil function and health (Li et al. 2017). Therefore, many researchers have recently focused on environment-friendly biomaterials such as Brassicaceae plant residues, which is considered a potential fumigant owing to its outstanding performance in disease control (Mazzola et al. 2017;Weerakoon et al. 2012). Brassica seed meals (BSMs) are rich in glucosinolates, which release isothiocyanates (ITCs) upon hydrolysis (Angus et al. 1994;Gimsing and Kirkegaard 2008). It has been widely reported that ITCs (especially allyl ITC) exhibit high biocidal activity against many soil-borne pathogenic bacteria, fungi, and nematodes (Handiseni et al. 2013;Kirkegaard et al. 1996;Mazzola et al. 2007Mazzola et al. , 2015Mazzola et al. , 2017. Soil fumigation with BSMs has been demonstrated to be effective in suppressing various plant diseases (Cohen et al. 2005;Mazzola et al. 2007Mazzola et al. , 2017; Ren et al. 2018;Wang and Mazzola 2019). In tomato, pepper, bean, and cabbage, BSMs have been reported to reduce the incidence of diseases caused by pathogenic bacteria and fungi such as Ralstonia solanacearum, Fusarium oxysporum f. sp., and Rhizoctonia solani (Abdallah et al. 2020;Chung et al. 2002;Ma et al. 2015;Pane et al. 2017). In addition, BSMs inhibit strawberry charcoal rot by reducing the pathogen (Macrophomina phaseolina) density in fumigated soil and root infection in strawberry rhizosphere soil (Mazzola et al. 2017). BSM treatments also effectively control apple replant disease by suppressing the plant pathogenic genera Phytophthora and Pythium and the nematode Pratylenchus penetrans in bulk soil and rhizosphere (Dupont et al. 2021;Mazzola et al. 2015;Weerakoon et al. 2012).
However, owing to the broad-spectrum biocidal activity of BSMs, BSM fumigation has an inhibitory effect not only on pathogens but also on most nontarget soil microorganisms (Dangi et al. 2017;Mazzola et al. 2001). The effects of BSM fumigation on various soil microorganisms leads to different quantitative changes in the soil microbiota (Mazzola et al. 2007;Reardon et al. 2013), which in turn result in changes in the soil microbial community composition (Fang et al. 2020;Mazzola et al. 2015). BSM fumigation affects the colonization of the rhizosphere by beneficial and harmful soil microorganisms (Martin 2003), which is strongly correlated with soil function and plant health. Some reports suggest that BSMs significantly affected the microbial community composition in bulk soil and rhizosphere, enriched disease-suppressing microorganisms, and improved the ability of soil microbiota to prevent reinfection by pathogens (Mazzola et al. 2015(Mazzola et al. , 2016. DuPont et al. (2021) showed that BSM application significantly altered composition of the bulk soil and rhizosphere microbiomes and reduced apple root density of pathogen Pythium ultimum. Ren et al. (2018) conducted pepper pot experiments to investigate the response of soil bacterial communities to BSMs amendment under the condition of Fusarium wilt infection, and found a significant increase in Actinobacteria-affiliated Streptomyces that may be involved in BSMinduced disease suppression. Although many studies have focused on the effects of changes in pathogen abundance and community composition following fumigant application, few studies have systematically and clearly elucidated the reconstruction of bacterial community structure after BSM incorporation into the soil. It remains poorly understood how soil bacterial communities rebuild following BSM fumigation and which microorganisms recolonize and thrive in fumigated bulk soil and rhizosphere soil.
It is well known that, with volatile biocidal compounds being released into the soil, fumigation is detrimental to soil bacterial diversity and community structure (Yan et al. 2015); however, soil microbial communities were found to be able to make a rapid recovery in terms of activity and diversity (Dangi et al. 2017;Dungan et al. 2006). The fumigation process and recovery process dominate the soil microbial community dynamics in the former and latter stages, respectively. However, little is known about the resistance of different microorganisms to fumigation extinction and their resilience after fumigation extinction. The specific mechanisms of microbial community reconstruction in fumigated soil and its effect on subsequent rhizosphere recruitment deserve more attention and further research. Therefore, unraveling the variation in bacterial community structure following fumigation extinction and recovery may contribute to a more comprehensive and in-depth understanding of reconstruction of bacterial community structure and effects of reconstruction processes on rhizosphere community assembly, which has significant implications for better management of fumigants to control soil-borne diseases.
We hypothesized that after fumigant application, soil bacterial community will initially undergo a broad-spectrum extinction with decreasing diversity and then a selective recovery process that determines the reconstruction of beneficial bacterial community structure in fumigated soil and subsequently affects the rhizosphere community composition and promotes plant health. On the basis of this hypothesis, we treated soil with fumigants to investigate the soil bacterial community composition after the extinction and recovery processes and examine their contribution to rhizosphere community assembly responsible for disease control. We aim to elucidate the mechanism of reconstruction of soil bacterial community structure and explain the variation in rhizosphere community composition after BSM fumigation. We focused on a) microorganisms selected by fumigation extinction and recovery following fumigation; b) and the contribution of fumigation extinction and recovery process to rhizosphere bacterial community assembly and pathogen suppression. An understanding of the bacterial community reconstruction mechanisms under BSM treatments will be vital to soil bio-fumigation management for soil-borne disease control in agriculture production systems.

Fumigants used in field experiments
Two types of Brassica seed meal (BSM), Brassica campestris seed meal (BcSM) and Brassica juncea seed meal (BjSM), were used for soil fumigation based on their outstanding inhibitory effects against soil-borne pathogen and plant disease in our previous research (Peng et al. 2021). These BSMs are the byproducts of oil extraction after seed cakes have been crushed and passed through a sieve (2 mm pore size). The carbon (C), nitrogen (N), phosphorus (P), and potassium (K) contents were 46.7%, 5.12%, 0.98%, and 1.21% in BcSM and 44.5%, 6.43%,1.25%, and 1.24% in BjSM, respectively. In addition, dazomet (C 5 H 10 N 2 S 2 , 98% purity; Wuhan Xingzhongcheng Technology Co., Ltd.) was used as a chemical fumigant for positive control (ML).

Site description and experiment design
The trial sites were located in Bozhou city, Anhui province, China. The generality of this experiment was assessed by conducting the field fumigation in two plastic greenhouses that had been used for two years for the continuous production of tomato (Solanum lycopersicum; greenhouse tomato [G-T]; 33°30′N 116°10′E) and cantaloupe (Cucumis melo var. saccharinus; greenhouse cantaloupe [G-C]; 33°45′N 115°40′E), respectively. The experiments began in June 2019 and ended in October 2019.
Before planting, fumigation was performed in June 2019 at the daytime temperature of 25-32 °C. BcSM and BjSM were added separately at a rate of 18 tons ha −1 . Dazomet was applied at a rate of 400 kg ha −1 following manufacturer's instructions. Soil without amendment was used as blank control (CK). Soil was incorporated with fumigants, then rotovated and mixed at the depth of 0-20 cm thoroughly. After application of fumigants, the soil was watered to 40%-60% moisture content. Then, the soil in all treatments was covered with 2 mm transparent plastic films to enhance fumigation effect, and the films were removed 2 weeks later. Four plots of 30-35 m 2 each were assigned to each treatment. 1 month after the film removal, 50 tomato and cantaloupe seedlings were, respectively, transplanted into each field plot under fumigation treatment. Then, 2 months after tomato and cantaloupe planting, the fruits were harvested and the disease incidence of plants in each plot was assessed by recording the number of healthy and wilted plants. The disease incidence was calculated using the following formula: Disease incidence = [number of wilted diseased plants/(number of wilted plants + number of healthy plants)] × 100%. Then, the disease control efficiency was calculated using the following formula: Disease control efficiency = [(disease incidence of the control treatment − disease incidence of the fumigation treatment)/(disease incidence of the control treatment)] × 100%. In addition, fruit weight was recorded individually, and the average was calculated.
The experimental plots were arranged in a randomized complete block design in both G-T and G-C. Control and fumigated plots received the uniform irrigation and fertility management in each greenhouse. After removing plastic films to terminate the fumigation, the chemical compound fertilizer of nitrogen (N), phosphorus (P) and potassium (K) and organic fertilizer were applied. The total amount of chemical compound fertilizer (N:P 2 O 5 :K 2 O, 15:15:15) was at a rate of 800 kg ha −1 and 1200 kg ha −1 for G-T and G-C, respectively, and the organic fertilizer with approximately 800 kg ha −1 and 3600 kg ha −1 were respectively supplied to G-T and G-C. During the plant growth period, weeding, flowering, and fruit thinning practices were carried out uniformly in all plots. Additionally, no additional fertilizer was used considering the nutrient enrichment of BSMs.

Soil sampling and analysis
At the time of sampling, a mixed soil sample was collected from the top 20 cm soil layer of the plot (four plots per fumigation treatment) using the checkerboard sampling method. During the experiment, in total, 52 soil samples were collected from each greenhouse in four periods, including T1: soil collected before fumigation; T2 (fumigation extinction): bulk soil collected immediately after plastic film was removed. The soils were fumigated for two weeks and microbial communities were extinct at T2; T3 (fumigation recovery): bulk soil collected 1 month after T2, at which microbiomes had naturally recovered for a month after fumigation ended; and T4: rhizosphere soil collected 2 months after T3 (plants had been growing for 2 months, and it was the time for harvest). Rhizosphere microorganisms are highly associated with plant diseases; therefore, at T4, rhizosphere soil was collected instead of bulk soil. All the soil samples were placed in an incubator with an ice bag and then transported and stored in a refrigerator at -20 °C for DNA extraction. In addition, the soil samples from T3 were stored at 4 °C for the analysis of soil chemical properties as follows: ammonia nitrogen (NH 4 + -N), nitrate nitrogen (NO 3 − -N), available phosphorus (AP), available potassium (AK), total nitrogen (TN), total phosphorus (TP), total potassium (TK), total organic carbon (TOC), and potential of hydrogen (pH).
High throughput sequencing and bioinformatics DNA was extracted from 0.5 g of each soil sample using Fast DNA Spin kit (MP Biomedicals, Shanghai, China) following the manufacturer's instructions. The V4 and V5 regions of the 16S rRNA gene were amplified using the primers 515F (5′-GTG CCA GCMGCC GCG GTAA-3′) and 907R (5′-CCG TCA ATT CCT TTG AGT TT-3′). The amplified PCR products were sequenced on the Illumina HiSeq platform. Then the sequences were analyzed using the QIIME software package (Caporaso et al. 2010). Low-quality sequences and noise were eliminated, and then the remaining sequences were clustered into operational taxonomic units (OTUs) with ≥ 97% similarity, and the taxonomic information was annotated using RDP classifier (Wang et al. 2007). After removing singleton OTUs, the number of OTUs and OTU taxonomy information were used for further analysis.
Analysis of variation in soil bacterial community richness, composition, and habitat niche breadth The significant differences in the soil variables and community richness (measured by observed OTUs) were calculated using the function "Analysis of variance (ANOVA)" in IBM SPSS 23.0 (IBM Corp., Armonk, NY, USA). In cases where the data failed the test for homogeneity of variance on the basis of the Duncan method, the data were analyzed using the Dunnett method (Kim 2014).
The principal coordinates analysis (PCoA), analysis of similarities (ANOSIM), and permutational analysis of variance (PERMANOVA) (Anderson and Walsh 2013) were conducted to examine the differences in community composition by using the "vegan" and "pairwiseAdonis" package (Martinez Arbizu 2020) in R 4.0.4 (R Core Team, Vienna, Austria). Mantel test was conducted to determine the correlation between predictive factors and community composition using the "vegan" package (Oksanen et al. 2020) in R. The "psych" package and "pheatmap" package (Galili et al. 2017) were used to calculate Pearson's correlation coefficients between environmental factors and bacterial taxon abundance. The Venn network were displayed and visualized to present the unique OTUs in each treatment by using Cytoscape 3.8.2 (Jiao et al. 2017). The "DESeq2" package in R (Love et al. 2014) was used to conduct a differential OTU abundance analysis between different communities by fitting a generalized linear model (GLM) based on the negative binomial distribution to normalization. The likelihood ratio test was used for significance testing of variance to identify significantly changed OTUs, and adjusted P-values (cutoff as 0.05) were calculated by Benjamini and Hochberg false discovery rate (FDR) for multiple testing. The "ggplot2" package in R was used for the graphical display of data.
Niche breadth calculated using the Levin's niche breadth index (B) equation (Pandit et al. 2009): where B j represents habitat niche breadth of OTU j in a metacommunity; N is the total number of communities in each metacommunity; P ij is the proportion of OTU j in metacommunity i . A high B indicates that the OTU occurs extensively and evenly along a wide range of locations, representing wide habitat niche breadth. Then, the average niche breadth of all OTUs within each sampling community ( B com ) were calculated as an indicator of habitat niche breadth at the community level. Bacterial communities with wider habit niche breadth are considered more widely distributed and more adaptable to environmental change (Peng et al. 2022).
Variance partitioning analysis and random forest analysis Variance partitioning analysis (VPA) was conducted to assess the influence of the former microbial community structure (T2) and environmental factors on the bacterial community composition in the postrecovery period (T3) using the "varpart" function in "vegan" package in R (Peres-Neto et al. 2006). Random forest (RF) analysis was performed to calculate the regression of genera (relative abundance > 0.05%) on one of the environmental factors (available phosphorus; AP) using the "randomForest" package in R (Edwards et al. 2018). The tenfold crossvalidation method was applied to determine the optimal assembly of genera correlated with nutrient (AP) content. The top 30 genera with the highest increase in mean square error of AP content were considered key microorganisms that responded strongly to nutrients in the recovery process after fumigation.

SourceTracker and species classification model
SourceTracker (Knights et al. 2011) was used to identify the putative sources of fumigated soil and rhizosphere microbes using "SourceTracker.r" package in R (Hartmann et al. 2016), to assess the effect of fumigation on the soil bacterial species pool during the fumigation process. The two putative sources of the target bacterial community under different treatments were assumed to be bacterial communities from unfumigated soil at the same sampling period and residual bacterial communities from the previous sampling period under the corresponding treatment.
For deciphering the contribution of species after fumigation to the recruitment of rhizosphere bacterial community, the species classification model (at the OTU level) was constructed on the basis of their behavioral characteristics (resistance and resilience level according to the changes in their abundance at T2 and T3, respectively) ( Figure S1). The "DESeq2" package in R was used to conduct a differential OTU abundance analysis between control and fumigated soil communities as described in 2.5. The species with significantly higher abundance in fumigated soil than in CK at the end of the fumigation period (T2) represented species equipped with higher resistance to fumigation extinction. The species with no significant change in abundance indicated moderate resistance. After the recovery of fumigated soil communities (T3), the resilience of species was determined by calculating the changes in their abundance. The species with high resilience indicated the increase in abundance owing to fumigation was still maintained after recovery (T3); the increase in abundance was caused only by the recovery process; or the decrease in abundance owing to fumigation was eliminated or reversed after recovery (T3). In addition, stable species indicated those that showed no significant changes in abundance after fumigation (T2) and recovery (T3). The species were categorized according to different levels of resistance and resilience and were identified to be enriched or depleted in the corresponding rhizosphere community, to determine the contribution of the resistance and resilience of these species to their recruitments to rhizosphere community. The species with average abundance > 0.005% were calculated in the model, and the abundance levels were compared using the "DESeq2" package in R.

Co-occurrence network analysis
Co-occurrence networks were constructed with the fumigated bulk soil (T3) and rhizosphere soil (T4) samples using the function "CoNet" in Cytoscape 3.8.2. The OTUs presented in at least two-thirds of all samples were chosen for calculation. Pairwise Spearman correlations between OTUs were calculated; the threshold value for the correlation coefficient was > 0.70 at a P value < 0.05. The main ecological clusters (modules) were identified and visualized using Gephi 0.9.2. The relative abundance of each module was obtained by averaging the relative abundance (transformed by z score) of all corresponding species in the module. The correlations between the abundance of pathogens and modules were calculated to determine the potential impact of ecological modules on the plant pathogens.

Results
Effect of fumigation on crop disease control and suppression of soil-borne pathogens Fumigation treatments significantly inhibited the wilting symptoms of tomato and cantaloupe (Table 1). In the control group without fumigants, the disease incidence of tomato and cantaloupe was 26.2% and 51.1%, respectively. BcSM, BjSM, and ML exhibited control efficiencies of 84.2%, 71.9%, and 33.9% in tomato greenhouse (G-T), respectively, and 49.3%, 54.7%, and 55.5% in cantaloupe greenhouse (G-C), respectively. In addition, BcSM and BjSM treatments showed significantly positive effects on cantaloupe yield, increasing single fruit weight to 2.06 and 1.91 kg, respectively (Table 1). These results showed that fumigation treatments significantly promoted plant health and growth and that BSMs used as a biofumigant were more effective than chemical fumigation with dazomet.
To evaluate the effect of bacterial community reconstruction after fumigation on disease control, co-occurrence networks were constructed using fumigated bulk soil (T3: after fumigation recovery) and rhizosphere soil (T4: during crop maturity) communities. Then, main ecological clusters of co-occurring OTUs were identified, and their role in the reduction of pathogen abundance was examined. Three Table 1 Plant disease incidence under different fumigation treatments Dissimilar lowercase letters in the same greenhouse soil in the same row represent a significant (P < 0.05) difference among the treatments. BcSM, Brassica campestris seed meal; BjSM, Brassica juncea seed meal; ML, dazomet; CK, control without fumigant application. "G-T" and "G-C" indicate greenhouses for tomato and cantaloupe, respectively and four main ecological clusters were chosen from the co-occurrence networks of bacterial communities in G-T and G-C, respectively (Fig. 1A). Module S1 and Module C1 showed the highest proportion of Firmicutes species among all modules, and its relative abundance under BSM fumigation groups was significantly higher than that in CK and ML groups, in both bulk and rhizosphere soil samples ( Fig. 1B and 1C). The species in Module S2 and Module C2 in the two greenhouses, respectively, mainly belonged to phyla Proteobacteria and Bacteroidetes, the relative abundance of which increased significantly in the rhizosphere, indicating that these microorganisms prefer the rhizosphere. In addition, BSM application resulted in a significant decline in the relative abundance in module S3 in G-T and Module C3 and Module C4 in G-C, which were mainly attributed to phyla Proteobacteria and Actinobacteria ( Fig. 1B and 1C). Fungal genus Fusarium was selected as a simplified biological indicator to assess the ecological significance of these modules in disease control. The correlation between the abundance of genus Fusarium and ecological modules was examined. The relative abundance values (Z score) in Module S2 and Module C2 showed a significantly positive correlation (r = 0.877 in G-T and 0.715 in G-C, P < 0.05) with the relative abundance of potentially pathogenic fungal genus Fusarium (Table 2), indicating that the rhizosphere recruitment was associated with a preference for pathogens. In contrast, the relative abundance of potentially pathogenic genus Fusarium was significantly negatively correlated with the relative abundance values in Module S1 and Module C1 (r = -0.525 in G-T and -0.745 in G-C, P < 0.05), indicating that these modules were associated with pathogen suppression ( Fig. 1B; Table 2).

Variation in soil bacterial community structure after fumigant application
To explain how disease-suppressive bacterial communities were restored after fumigant application, we mainly focused on the variation in soil bacterial community structure. The distribution of time and fumigation treatments was well differentiated on the principal component axis, which together explained 58.51% and 64.58% of the total community variance in G-T and G-C, respectively ( Fig. 2A). Notably, the results of significance testing of variance suggested that fumigant application significantly affected bacterial community composition and structure (Table S1, P < 0.05), and this effect decreased over time (Fig. 2B). Moreover, during the restoration of soil bacterial community structure in fumigated soil, Firmicutes and Proteobacteria were mainly responsible for changes in community composition at the phylum level (Fig. 2C). In addition, BSMs showed greater effects than dazomet (ML) on the bacterial community composition throughout the process, as indicated by the higher levels of community variation ( Fig. 2B; Table S1).
For deciphering the restoration of bacterial communities in fumigated soil, this study investigated bacterial community changes during two key processes: fumigation extinction and recovery. After fumigation extinction (T2), the soil bacterial communities in the two greenhouses showed lower habitat niche breadth and species richness than control (Figure S2). However, the phylum Firmicutes exhibited higher species richness after fumigation; in contrast, the other main phyla (top eight in relative abundance) showed significant reduction in species richness (Table S2, P < 0.05). Moreover, Firmicutes species dominated the common fumigation-unique species among the three fumigation treatment groups (Fig. 3), suggesting its higher tolerance to fumigation. In contrast, although Proteobacteria also exhibited a large number of fumigation-unique species in the fumigation treatments, most of these species were present independently in each fumigation treatment, indicating the unreliability and variability of Proteobacteria in resistance to fumigation. In addition, volcano plots showed that the species significantly enriched under fumigation treatments were dominated by members of Firmicutes (Fig. 4A).
One month after fumigation ended (T3), bacterial communities showed significant recovery with the restoration of habitat niche breadth and species richness ( Figure S2). Although the species richness of the main phyla (except for Firmicutes) significantly increased after recovery (T3), the species number for these phyla in the BSM treatment groups remained significantly lower than that in the CK group (Table S2, P < 0.05). In the BcSM and BjSM treatment groups of G-T, 169 and 167 species were significantly enriched but 149 and 168 species were significantly depleted, respectively; in the BcSM and BjSM treatment groups of G-C, 168 and 225 species were significantly enriched, while 63 and 83 were significantly depleted, respectively (Fig. 4A). After classifying the significantly altered species into corresponding phyla, the column chart revealed that Proteobacteria, Actinobacteria, and Firmicutes were the most dominant phyla in the increased species under BSMs treatments (Fig. 4B). Moreover, although ML also significantly affected soil bacterial communities, it affected species richness and community composition to a lesser extent than BSMs ( Fig. 4; Figure S2B; Table S1).
Factors affecting the restoration of bacterial community structure after fumigant application To investigate the mechanism of restoration of soil bacterial community structure after fumigation extinction, we focused on the effects of bacterial communities before recovery (T2) and environmental factors on the variation in bacterial community composition after recovery (T3). The variance partitioning analysis (VPA) showed that the initial state of community recovery (T2) and environmental factors both significantly influenced bacterial community recovery (T3) and respectively explained 19% and 11% of the variance and 17% and 7% of the variance in bacterial community composition after recovery in G-T and G-C, respectively (Fig. 5A). The linear regression models showed a positive correlation between the changes in community composition at T2 (before recovery) and T3 (after recovery; R 2 = 0.662, P < 0.001 for G-T; R 2 = 0.460, P < 0.001 for G-C) (Fig. 5B). Among environmental factors, AP was the most important environmental factor, showing the highest positive linear correlation (R 2 = 0.442, P < 0.001 for G-T; R 2 = 0.617, P < 0.001 for G-C) with variation in bacterial communities (Fig. 5B). To decipher the key microorganisms that respond to nutrient content during community recovery after fumigation treatment, RF analysis was performed to identify 30 genera most highly associated with nutrient (AP) content. Most of the identified genera belonged to the phyla Proteobacteria and Actinobacteria, indicating the advantages and key predictive roles of these two phyla in the recovery process of bacterial community ( Figure S3). In addition, heatmaps showed that the relative abundance of the identified genera in BSM (BcSM and BjSM) treatment groups was significantly different from that in the CK and ML groups, indicating that BSMs, owing to their nutrient supply, caused more changes in the bacterial community composition than dazomet in the recovery process ( Figure S3). Moreover, BSMs significantly enriched the content of soil nutrients other than AP, including total organic carbon (TOC), total nitrogen (TN), and total phosphorus (TP; Table S3, P < 0.05). Moreover, these nutrients showed a significant positive correlation with the genera Streptomyces, Actinomadura, Micromonospora, and Saccharomonospora belonging to Actinobacteria, the genera Luteimonas and Pseudoxanthomonas belonging to Proteobacteria, and the genera Bacillus and Hazenella belonging to Firmicutes in fumigated soil after recovery in G-T and G-C (Fig. 6), indicating that these three phyla had proliferation advantages during the recovery of bacterial communities under BSM treatments with nutrient supply.

Effect of reconstruction processes on rhizosphere microbial recruitment under fumigation treatment
To decipher the mechanisms of bacterial community reconstruction in fumigated soil and subsequent rhizosphere microbial recruitment, microorganisms that changed during different periods (T2: after fumigation extinction, T3: after recovery, and T4: during crop maturity) under fumigation treatments were tracked using the SourceTracker analysis and species classification model (based on species resistance to fumigation and resilience after fumigation).
When rhizosphere soil samples were used (T4), the differences in the bacterial community composition between the control and fumigated groups decreased but remained significant ( Fig. 2B; Table S1, P < 0.05). The SourceTracker results revealed that bacterial communities in fumigated soil (T3) under the BcSM, BjSM, and ML treatments respectively constituted 42.5%, 52.0%, and 30.5% of the corresponding rhizosphere communities in G-T and 30.8%, 35.8%, and 16.3% of the corresponding rhizosphere communities in G-C (Fig. 7A), indicating that Fig. 1 Co-occurrence network analysis revealing ecological modules in fumigated soil and rhizosphere communities. A Network diagram with nodes colored according to each of the main ecological clusters for tomato greenhouse (G-T) and cantaloupe greenhouse (G-C); B Relative abundance (Z score) of ecological clusters in different fumigation treatment groups in G-T and G-C; C Operational taxonomic unit (OTU) number properties of the dominant phyla in the main ecological clusters in G-T and G-C. Different letters represent significant (P < 0.05) differences among the treatments in bulk soil (A, B, C) and rhizosphere soil (a, b, c). BcSM, Brassica campestris seed meal; BjSM, Brassica juncea seed meal; ML, dazomet; CK, control without fumigant. "T3" represents the sampling time one month after fumigation extinction, at which the fumigated soil communities have naturally recovered and reconstructed. Crops seedlings were also planted at this time; "T4" represents the sampling time 2 months after T3, and it was the time for harvest ◂ fumigation significantly affected the bacterial species pool in the rhizosphere (T4), and fumigation with BSMs had a greater effect on rhizosphere species sources than fumigation with dazomet.
In tomato rhizosphere community (G-T), 154, 200, and 102 species were significantly enriched, whereas 167, 239, and 90 species were significantly depleted under BcSM, BjSM, and ML treatments, respectively. In cantaloupe rhizosphere community (G-C), 70, 66, and 13 species were significantly enriched, whereas 36, 60, and 10 species were significantly depleted under the BcSM, BjSM, and ML treatments, respectively (Fig. 7B). To evaluate the effect of microbial selection by fumigation on rhizosphere recruitment, the species classification model was used to identify species with different levels of resistance and resilience after fumigation extinction (T2) and recovery (T3), respectively. It showed that 136 and 60 species were enriched and depleted after fumigation extinction (T2) in BcSM of G-T, respectively, which were considered as high-resistance and low-resilience species, respectively. Then after recovery (T3), of the 136 species with high resistance, 42 and 94 species had high and low resilience, respectively; of the 60 species with low resistance, 40 and 20 species had high and low resilience, respectively (Fig. 7B). The contribution of fumigation to rhizosphere recruitments was evaluated by tracking the resistance and resilience levels of rhizosphere species that were significantly altered under fumigation treatments. The species classification model showed that, of the enriched species in tomato rhizosphere communities under BcSM and BjSM treatments, 12 and 13 species showed high resistance, but only 0 and 3 species showed low resistance, respectively. Conversely, of the significantly depleted species in the tomato rhizosphere communities under BcSM and BjSM treatments, 23 and 20 were low-resistance species, whereas only 6 and 6 were high-resistance species, respectively ( Figure S4A). Similar results were observed in cantaloupe rhizosphere communities (G-C) under BcSM and BjSM treatments, suggesting that high-resistance species were more likely to be enriched in the rhizosphere, while low-resistance species were easier to be depleted. Moreover, species resilience also affected the species assembly in the rhizosphere under BSM treatments ( Fig. 7B; Figure S4A). Among the significantly enriched species in tomato rhizosphere communities (G-T) under BcSM and BjSM treatments, there were 78 and 76 highresilience species but only 0 and 3 low-resilience species, respectively. In contrast, of the significantly depleted species in tomato rhizosphere communities under BcSM and BjSM treatments, 89 and 96 species showed low resilience, and 8 and 6 species showed high resilience, respectively. Similar results were also presented in the cantaloupe rhizosphere communities (G-C). These results suggested that microbial resistance and resilience influence their recruitment to the rhizosphere under the BSM treatments. In addition, taxonomic analysis revealed that the species with high resistance were largely classified into the phylum Firmicutes, and the species with high resilience were mainly classified into the phyla Proteobacteria, Actinobacteria, and Firmicutes ( Figure S4B), indicating that these phyla play key roles in the bacterial community reconstruction following BSMs application. However, species resistance and resilience were not observed to contribute to the variance in rhizosphere community composition in the ML group.

Discussion
Firmicutes and Proteobacteria are key members responsible for the community reconstruction under BSM treatments Fumigation treatments significantly promoted plant health and BSM treatments enriched potentially beneficial bacteria to control pathogens after bacterial community reconstruction ( Fig. 1; Table 1 and 2), indicating the presence of beneficial communities after BSMs application, as described in previous studies (Dupont et al. 2021;Mazzola et al. 2016;Ren et al. 2018). Therefore, in the present study, we separated the bacterial community reconstruction during the fumigation process into two stages-fumigation extinction and subsequent recovery-aiming to systematically decipher the mechanisms of community reconstruction and rhizosphere recruitment following fumigation treatment. After fumigation extinction, soil bacterial communities showed lower habitat niche breadth and species richness ( Figure S2), indicating that fumigation created unfavorable habitats and limited the survivability of community members. Biocides released into the soil by fumigants eliminate not only pathogenic but also non-pathogenic microorganisms, which in turn reduces soil bacterial community diversity and activity (Dangi et al. 2017;Mazzola et al. 2001). Although different microorganisms respond differently to fumigants (Kirkegaard et al. 1996), a biocidal preference for certain microbes has been observed in the extinction process. A previous study reported that the relative abundance of Firmicutes increased markedly after fumigation with chloroform (Domínguez-Mendoza et al. 2014); similarly, in our study, Firmicutes showed significant higher species richness and relative abundance after fumigation extinction and showed a higher number of unique and enriched species than other phyla under fumigation treatments ( Fig. 2C and 3 and 4A; Table S2), indicating that Firmicutes had a higher resistance to fumigation than other phyla. The superior resistance of Firmicutes to fumigation may be attributed to its special physiological characteristics. Firmicutes resist adverse environmental conditions including heat, radiation, and chemicals by employing the survival strategy of forming endospores (Beskrovnaya et al. 2021;Filippidou et al. 2016;Mandic-Mulec and Prosser 2011), which are considered the most resistant cellular structures (Abecasis et al. 2013). In addition, Firmicutes use other strategies such as exogenous DNA uptake, chemotaxis, motility, biofilm formation, and functional diversification to adapt to extreme environmental conditions (Mandic-Mulec and Prosser 2011). Thus, the fumigated environments rich in biocidal gases may provide Firmicutes species a higher survival advantage than other microorganisms with weaker resistance, leading to the dominance of Firmicutes and the depletion of other phyla in soil bacterial communities after fumigation (Fig. 2C). Proteobacteria exhibited the highest number of unique species under each fumigation treatment (Fig. 3), likely owing to their variable morphology and versatile physiology (Shin et al. 2015), indicating that Proteobacteria species also adapt to adverse conditions. After removing the sealing film, the biocidal vapor in soil environment was released and the habitat of soil microorganisms gradually returned to normal, leading to a significant increase in habitat niche breadth and community diversity ( Figure S2). Simultaneously, the variation in soil bacterial communities caused by fumigation was significantly reduced ( Fig. 2B; Table S1). However, there remained significant differences between microbial communities in fumigated and unfumigated soil, which can be partly explained by the effects of fumigation extinction and environmental factors (Fig. 5). The history and legacy of microbial community structure determines the microbial community restoration following disruption on a large scale (Hawkes and Keitt 2015;Jurburg et al. 2017). This pattern explains the significant positive correlations between the changes in microbial community composition in the postfumigation period (T2) and post-recovery period (T3). In addition to the legacy effect of fumigation extinction, another influencing factor of the changes in community composition in fumigated soil are environmental factors (Fig. 5), which play important roles in determining the recovery of microbiomes following habitat disruption (Xiang et al. 2014). Soil nutrients, especially carbon, nitrogen, and phosphorous, strongly influence soil microbial biomass, diversity, and community composition (Liu et al. 2012;Smith and Prairie 2004;Su et al. 2015;Wang et al. 2014). BSMs enrich soil nutrients by providing nitrogen, phosphorus, and TOC (Table S4) (Mazzola et al. 2007;Reardon et al. 2013). The content of nutrients was significantly correlated with the community composition after fumigation recovery (Fig. 5B), which partly explains why treatment with BSMs resulted in a greater variance in community composition after the recovery period than treatment with nutrient-free ML (Table S1). Among the nutrients, phosphorus is least available in the soil but is an essential macronutrient for microbes (Raghothama  1999). When soil carbon is sufficient, the P deficiency directly limits soil bacterial growth (Demoling et al. 2007); therefore, considering the dynamic microbial growth in the soil environment following fumigation extinction, available phosphorus (AP) may be the most important factor in determining community recovery (Fig. 5B).
In the presence of nutrients provided by BSMs, some members of Proteobacteria, Actinobacteria, and Firmicutes increased significantly after fumigation recovery, resulting in significant changes in community composition, which were significantly different from those in the ML treatment ( Fig. 4B and S3). Genera belonging to Proteobacteria, Actinobacteria, and Firmicutes, including Streptomyces, Actinomadura, Micromonospora, Saccharomonospora, Luteimonas, Pseudoxanthomonas, Bacillus, and Hazenella, exhibited a significant positive correlation with the nutrient content of BSMs (Fig. 6), suggesting that they had an advantage in nutrient utilization during community recovery, likely owing to their physiological properties. For example, the members of Proteobacteria exhibit a wide spectrum of resource utilization and advantages in growth (Llado and Baldrian 2017). Moreover, copiotrophs such as Proteobacteria and Actinobacteria have faster growth rates and acquisition of nutrients rates than oligotrophs, allowing them to be highly competitive for resources and thus leading to their preferential augmentation in environments with high levels of nutrients Fierer et al. 2007;Gao and Wu 2018;Ho et al. 2017;Orwin et al. 2018). In fumigated soil where residual microorganisms share abundant resources for survival, the enrichment of soil nutrients and dead microbes leads to the dominance of copiotrophic members in bacterial communities (Fierer et al. 2012;Orwin et al. 2018).
The microbial selection by the fumigated environment determined the reconstruction of disease-suppressive bacterial communities Firmicutes and Proteobacteria represent highly resistant and resilient microorganisms in bacterial communities in fumigated soil ( Figure S4); they are the key microbes for community reconstruction during the processes of fumigation extinction and recovery, respectively. Rhizosphere microbial community composition is determined not only by plant root zone but also by bulk soil microbial community composition (de Ridder-Duine et al. 2005;Marschner et al. 2001). The fumigated bulk soil bacterial communities are crucial for the source and assembly of rhizosphere soil microbial community. Moreover, highly-resistant microorganisms may be better able to cope with unfavorable metabolites from hosts and other microorganisms during rhizosphere filtration; while highly-resilient microorganisms may have more advantages in resource utilization to occupy more ecological niches in rhizosphere environments. Our results showed that microorganisms with higher resistance or higher resilience are more likely to accumulate in the rhizosphere ( Fig. 7 and S4), emphasizing the effect of fumigation on the preference for certain microbes in rhizosphere recruitment under fumigation treatments, indicating that microbial resistance and resilience influence their recruitment to the rhizosphere.
Therefore, the restoration of beneficial bacterial communities may be derived from the microbial selection by the fumigated environment on the basis of physiological properties. In a fumigated soil environment, gram-positive bacteria are less affected by biocidal agents than gramnegative bacteria owing to their different cell Fig. 2 Bacterial community variation in different periods during fumigation and subsequent plant growth. A. Principal coordinate analysis (PCoA) of bacterial communities based on Bray-Curtis distance; B. Bray-Curtis distance of bacterial communities between the fumigation treatments and the control (CK) treatment; C. Bacterial community composition at phylum level. BcSM, Brassica campestris seed meal; BjSM, Brassica juncea seed meal; ML, dazomet; CK, control without fumigant application. "T1" represents the sampling time before fumigation; "T2" represents the sampling time at the end of the fumigation. The soils were fumigated for two weeks and microbial communities were extinct at T2; "T3" represents the sampling time one month after T2, at which the fumigated soil communities have naturally recovered and reconstructed. Crops seedlings were also planted at this time; "T4" represents the sampling time 2 months after T3, and it was the time for harvest. Bulk soils were sampled at T1, T2, and T3, while rhizosphere soils were sampled at T4. "G-T" and "G-C" indicate greenhouses for tomato and cantaloupe, respectively ◂ wall structure and their ability of spore formation (Dangi et al. 2017;Zelles et al. 1997). Therefore, fumigant application may lead to an increase in the proportion of gram-positive members (such as Firmicutes and Actinobacteria) in soil bacterial communities. Notably, the colonization  (Emmert and Handelsman 1999;Lee et al. 2021;Mendes et al. 2011). Our results showed that Module S1 (in G-T) and Module C1 (in G-C) that enriched Firmicutes in the fumigated soil and rhizosphere community had significant negative effects on the potentially pathogenic genus Fusarium (Table 2); however, assessing multiple pathogens (rather than one) may be more representative for potentially pathogenic soil microbes in complex greenhouse soils. Moreover, BSM nutrients significantly enriched the gram-positive genera Bacillus and Streptomyces (Fig. 6), which have been widely reported to have significant potential in suppressing soil pathogens and promoting plant growth by producing metabolites (Babalola 2010;Ek-Ramos et al. 2019;Huang et al. 2019;Khabbaz et al. 2015;Perez-Garcia et al. 2011). Therefore, in the present study, the prevalence of Firmicutes and Actinobacteria members after fumigation with BSMs may be a biomarker of effective microbial community reconstruction. On the other hand, the contribution to pathogen suppression by Proteobacteria members enriched in Module S1 (in G-T) and Module C1 (in G-C) suggested that the role of some gram-negative bacteria in disease prevention should not be overlooked. In the highly-resilient phylum Proteobacteria, microorganisms enriched after fumigation recovery may have high productivity and strong resource competition, which can defense against pathogen invasion by occupying more ecological niches. In contrast, most gram-negative pathogens without the ability to form spores are considered to not survive well in the soil (Raaijmakers et al. 2009) because soil pathogens need to infect host plants to acquire nutrients and they colonize and multiply in favorable habitats (Fatima and Senthil-Kumar 2015;Yuan et al. 2018). Moreover, BSMs have an inhibitory effect on the virulence of the pathogens, which plays an important role in survival in soil and host plant infection (Peng et al. 2021), indirectly inhibiting the proliferation of pathogens. Therefore, in fumigated soils before planting, pathogens may have a lower advantage in recovery and proliferation compared with nonpathogenic microbes, which may also be responsible for the emergence of healthy microbiomes after BSM application.
Compared with the single-ingredient chemical fumigant dazomet, BSMs had greater effects on the soil bacterial community diversity and composition after the fumigation extinction (T2) and recovery (T3) ( Fig. 2; Fig. 4; Table S1 and S2), which may be due to the diversification of antibacterial ingredients and enrichment of nutrients in BSMs. BSMs contain various antibiotics such as allyl ITC, 3-butenyl glucosinolate, and 2-phenylethyl ITC (Mazzola et al. 2017;Peng et al. 2021;Smolinska et al. 1997a, b), which provided a more broad-spectrum inhibitory effect against soil microorganisms than that of observed with dazomet. Moreover, the enrichment of soil nutrients by BSMs resulted in higher community variability than that observed with dazomet due to the effect of nutrient levels on the community composition during the post-fumigation recovery. Similar to previous reports that BSMs promotes the recruitment of beneficial bacterial communities and suppresses pathogens in the rhizosphere (Mazzola et al. 2015;Ren et al. 2018), the present study showed that the nutrient-rich BSMs provide more sustainable disease suppression by recruiting beneficial microorganisms and enriching potential disease-suppressing ecological clusters ( Fig. 1; Table 2). Compared with chemical fumigants, which are detrimental to soil microbial diversity and ecological functions (Cheng et al. 2020), nutrient-rich BSMs enhance soil properties and improve microbial activity in the recovery process following fumigation. Moreover, the application of bioorganic fertilizers following fumigation may enrich potentially Fig. 3 Venn network depicting the unique species in each fumigation treatment and the control at the end of the fumigation extinction. The small circular nodes in the figure represent different unique OTUs, and the different colors indicate that they belong to distinct phyla; the three large diamond, triangle, and arrow nodes represent the fumigation treatments of Brassica campestris seed meal (BcSM), Brassica juncea seed meal (BjSM), and dazomet (ML), respectively, and the square node indicates the control without fumigant application (CK); the small nodes are connected to the large nodes by lines, indicating that the unique OTUs are observed in the corresponding treatments. "G-T" and "G-C" indicate greenhouses for tomato and cantaloupe, respectively beneficial microorganisms and effectively suppress pathogens, leading to sustainable disease suppression (Deng et al. 2020;Zhao et al. 2017). Thus, BSMs may provide greater opportunities than dazomet to restore beneficial bacterial communities for the control of soil-borne plant diseases. Nevertheless, appropriate management practices during the fumigation processes are still required to promote the soil healthy. For example, the supplement of trace elements such as iron should also be considered in community restoration after fumigation, which potentially affect the interspecific relationships of microbial communities and the occurrence of soil-borne disease (Aznar et al. 2015;Verbon et al. 2017). In addition, reduction of chemical fertilizers is one of alternatives after the enrichment of soil nutrients by BSMs, as excessive chemical fertilizer inputs may lead to soil salinization and acidification .
In general, the mechanism of bacterial community reconstruction in fumigated soil is associated with a preference for microorganisms with high resistance and resilience to fumigation, which is helpful in understanding the generation of beneficial bacterial communities and worth considering in practical fumigant applications. In our study, gram-positive bacteria (especially Firmicutes) showed higher tolerance to fumigation than gram-negative bacteria; thus, fumigation may be more effective in gram-negative pathogens than in gram-positive pathogens. Moreover, copiotrophic bacteria (especially Proteobacteria) have greater advantages in the recovery of microbial communities in fumigated soil. Firmicutes and Proteobacteria members with high resistance and resilience screened by fumigation are better able to resist unfavorable environments and utilize resources, thus potentially thrive and occupy stable ecological niches in complex rhizosphere environments with intense resource competition, thereby inhibiting pathogen invasion and disease development. Our study provides researchers with a fundamental ecological perspective for understanding mechanisms of reorganization of microbial communities in fumigated soil systems; however, the specific taxonomic classification and physiological properties involved in microbial responses to soil fumigation require further research and emphasis in fumigant application and soil management. In addition, although 16S amplicon sequencing data is sufficient to evaluate the differences in resistance and resilience among microorganisms by comparing their relative abundance, the measurement of microbial absolute abundance or biomass would help us better track actual changes in the soil microbial population and reveal the resistance and resilience of microorganisms after fumigant application. Finally, although bio-fumigation is effective in improving diseased soils, the primary choice for agriculture production systems should be to identify the causes of soil-borne pathogen problems (monocultures, decreasing soil biodiversity, high-input industrial agriculture, etc.) and establish rational soil management systems instead of fixing problems with end-of-pipe solutions.

Conclusion
BcSM, BjSM, and dazomet fumigation significantly changed the soil bacterial community composition in the extinction and recovery stages. Moreover, compared to dazomet, BSMs had a greater impact on the diversity and composition of fumigated bulk soil and rhizosphere bacterial communities because of its greater biocidal activity and richer nutrient supply. Fig. 4 Volcano plots depicting the changes in bacterial communities in fumigated soil at the operational taxonomic unit (OTU) level. A Volcano plot showing the enriched and depleted OTUs after the fumigation extinction (T2); B volcano plot showing the enriched and depleted OTUs after the community recovery (T3). In volcano plots, the horizontal axis represents the average normalized OTU abundance, and the vertical axis represents the fold change. Red and blue circles indicate OTUs that increased and decreased significantly (fold change > 2 or < 0.5 and P-value < 0.05) in the fumigation treatments (BcSM, BjSM, and ML) compared with the control without fumigant application (CK), respectively. Histogram shows the number of significantly altered OTUs in different phyla. BcSM, Brassica campestris seed meal; BjSM, Brassica juncea seed meal; ML, dazomet; CK, control without fumigant application. "T2" represents the sampling time at the end of the fumigation. The soils were fumigated for two weeks and microbial communities were extinct at T2; "T3" represents the sampling time one month after T2, at which the fumigated soil communities have naturally recovered and reconstructed. "G-T" and "G-C" indicate greenhouses for tomato and cantaloupe, respectively

G-C
Gram-positive bacteria (especially Firmicutes) showed significantly higher resistance to extinction following fumigation. In the recovery stage following fumigation, the initial condition of microbial community and soil nutrient content (especially AP) determined the soil bacterial community diversity. Copiotrophic bacteria such as Proteobacteria and Actinobacteria showed higher resilience in the recovery stage following fumigation. In addition, the nutrients supplied by BSMs enriched the potentially beneficial genera Bacillus and Streptomyces. The advantage of soil microorganisms with high resistance and resilience (especially Firmicutes and Proteobacteria) during the stages of fumigation and recovery could be partly transmitted to the rhizosphere community; in other words, the resistance and resilience of microorganisms facilitate their recruitment to the rhizosphere, thus potentially occupying more ecological niches and preventing pathogen invasion and disease development. Compared with the chemical fumigant dazomet, BSMs facilitated the recruitment of beneficial microorganisms to the rhizosphere community with a higher potential to inhibit pathogens. Fig. 5 Effect of environmental factors on the variation in soil bacterial communities after the recovery period following fumigation. A Variance partitioning analysis (VPA) to assess the influence of microbial community composition at the end of the fumigation extinction (T2) and environmental factors on the bacterial community composition after the recovery period. B Linear regression to determine the correlation between the bacterial community composition after recovery (T3) and the following factors: NH 4 + -N, ammonia nitrogen; NO 3 − -N, nitrate nitrogen; AP, available phosphorus; AK, available potassium; TN, total nitrogen; TP, total phosphorus; TK, total potassium; TOC, total organic carbon; pH; and the bacterial community composition at the end of fumigation extinction (T2). The red line represents a significant correlation (P < 0.05), and the blue line represents a non-significant correlation (P > 0.05). "G-T" and "G-C" indicate greenhouses for tomato and cantaloupe, respectively ◂ Fig. 6 Heatmaps showing the correlation (Pearson coefficient) between the relative abundance of the top 30 genera and the content of soil environmental variables. The red blocks represent positive correlation, and the blue blocks represent negative correlation. "*", "**", and "***" represent significant correlations with a P value less than 0.05, 0.01, and 0.001, respectively. NH 4 + -N, ammonia nitrogen; NO 3 − -N, nitrate nitrogen; AP, available phosphorus; AK, available potassium; TN, total nitrogen; TP, total phosphorus; TK, total potassium; TOC, total organic carbon. "G-T" and "G-C" indicate greenhouses for tomato and cantaloupe, respectively Author contributions Jiangang Li: Conceived the study; Junwei Peng, Minchong Shen and Ruihuan Chen: Field experiments design and perform; Junwei Peng and Hong Liu: Collected the data; Junwei Peng and Yang Sun: analyzed the data; Junwei Peng, Qin Liu, Jiangang Li, and Yuanhua Dong: Led the writing and revision of the manuscript. All authors contributed critically to the drafts and gave final approval for publication.

Data availability
The raw sequences were deposited in NCBI SRA database (BioProject accession no. PRJNA870165). The processed data and code used in this study are available at https:// github. com/ reffi sh/ BSM-fumig ation. git.

Declarations
Competing interests The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. Fig. 7 Concept diagram of the mechanisms of bacterial community reconstruction in fumigated soil and subsequent recruitment in rhizosphere on the basis of operational taxonomic unit (OTU) changes. A SourceTracker analysis to determine the putative sources of bacterial communities in the different sampling time after fumigation. B Species classification model to evaluate the contribution of bacterial post-fumigating resistance and resilience ability to their subsequent colonization in rhizosphere. In panel A, it was assumed that the bacterial community under fumigation treatment was divided into two parts: the first part was the bacterial community unaffected by fumigation, which is considered to be from native soil communities (CK); the second part was the community affected by fumigation, which was left over from the previous stage. Green curve showed the percentage of first part, the red line indicated the percentage of second part, and the remaining part (unlabeled) indicates percentage of unknown sources. In panel B, soil bacterial species were classified into highly resistance, medium resistance, and low resistance based on their abundance changes in soil after fumigation (T2), accordingly, the resilience levels were then further classified by their abundance changes after recovery (T3). Each species can be classified as a certain level of resistance or resilience, which may determine their colonization in the rhizosphere. In rhizosphere communities (T4), resistance and resilience levels of species significantly altered under fumigation treatments were tracked to assess the effect of fumigation extinction and recovery processes on rhizosphere recruitments, respectively. The red symbols " + " and "-" represent species with relatively high and low resistance to fumigation; the red symbol "O" represents the species with a moderate level of resistance to fumigation; "↑" and "↓" represent species with relatively high and low resilience after fumigation; " = " represents species with stable relative abundance. The numbers in the symbols represent the number of species corresponding to the resistance and resistance levels. The histogram in the species classification model indicates the number of enriched or depleted species in the rhizosphere under each fumigation treatment classified on the basis of resistance and resilience levels. BcSM, Brassica campestris seed meal; BjSM, Brassica juncea seed meal; ML, dazomet; CK, control without fumigant application. "T2" represents the sampling time at the end of the fumigation. The soils were fumigated for two weeks and microbial communities were extinct at T2; "T3" represents the sampling time one month after T2, at which the fumigated soil communities have naturally recovered and reconstructed. Crops seedlings were also planted at this time; "T4" represents the sampling time 2 months after T3, and it was the time for harvest. Bulk soils were sampled at T2 and T3, while rhizosphere soils were sampled at T4. "G-T" and "G-C" indicate greenhouses for tomato and cantaloupe, respectively ◂