The field test plot is sited in the coastal area of Lianyun New-Town, Lianyungang City, Jiangsu Province (E119°22′, N34°76′). The climate is of the warm temperate type, with a sub-tropical transitional humid monsoon climate (http://www.weather.com.cn/cityintro/101191001.shtml); the soil is of typical CSS type. The CSS in this plot is formed by hydraulically transported and filled by dredger and mud pump from the bottom of beach in coastal zone in the process of reclamation project (An et al., 2022), with characteristically high salinity, viscosity and moisture content.

From July to August 2020, the soil of the test plot was improved by adding exogenous substances and planting salt-tolerant plants. The implementation scheme of CSS improvement by adding different exogenous substances is displayed in Table 1 , in which three repetitions are set for each sample plot (2 m * 4 m). During the implementation of the scheme, the first 200 mm depth of soil on the surface layer of the sample plot was first ploughed, then the exogenous additives were added before being ploughed again, evenly mixing the additives and soil. Among them, the organic fertilizer is provided by Lianyungang Golden Coast Development and Construction Co. Ltd, Lianyungang City, and the nutrient composition is found to be as follows: TOC 3.19%, total nitrogen (TN) 28.220 g·kg^-1, alkali-hydro nitrogen 4.865 g·kg^-1, available potassium 1.035 g·kg^-1, available phosphorus 2.315 g·kg^-1, moisture content 15.14%, pH 6.77; the sand is local river-sand of Lianyungang city, with the grain diameter ranging between 1 and 2.5 mm; mineral powder of serpentine was purchased from the serpentine mining area in Donghai County, Lianyungang City, with the grain diameter ranging between 1 and 2.5 mm. In addition, the salt-tolerant plants S. cannabina, Medicago sativa, Limonium bicolor, and Amaranthus tricolor were planted in the untreated CSS of the sample plot for two consecutive years previously, of which only S. cannabina could grow well and form a single vegetative community. Based on this, S. cannabina was used as an experimental species in CSS treated with different exogenous substances, and the planting density was about 240 plants·(m²)^-1.

From October to November 2020, rhizosphere soil and soil around the roots of S. cannabina and the bare soil in four sample plots were collected, with three replicates for each group of samples. Meanwhile, for each repetition of the sample, the soil with the same weight of the five adjacent sampling points was selected and mixed evenly. Specific sampling methods include: vertically pulling out the plants, shaking off the bulk soil attached to the roots, and the soil attached to the surface of the root of the S. cannabina was collected in a sterile sampling bag with a sterilized brush, which is the rhizosphere soil of S. cannabina, for total DNA extraction and high-throughput sequencing; the sterilized shovel was adopted to collect the soil around the S. cannabina roots within a 100-mm radius centered on the root and the bare soil without plant growth, forming samples of the soil around the roots of S. cannabina and the bare soil respectively, for determination of soil physico-chemical properties; in addition, a 100 cm³ ring knife was utilized to collect the soil around the S. cannabina roots within 100 mm, for determination of porosity. The depth of all the collected soil samples ranged between 50 and 150 mm. The collected samples were stored and transported in dry ice, and frozen at -80 °C. At the same time, the whole plants of S. cannabina grown on the ground of four sample plots were collected, and each group of plant samples contained six replicates. The dry biomass of the plants was measured after drying (at 45°C).

The soil around the roots were ground (to pass through a 60-mesh sieve) after air drying, and the detection methods of soil physico-chemical properties included: soil mineral crystal composition was determined by using an X-ray crystal diffractometer (Olympus, BTX-526, USA); total salt was measured by weighing deionized soil suspension with a soil to water ratio of 1:5 (w/v) (Lv and Li, 2010); electrical conductivity (EC) was identified by conductivity meter (SX731, Shanghai Sanxin, China) using deionized soil suspension with a soil water ratio of 1:2 (w/v); the cation exchange capacity was measured by referring to Chinese National Standard NY/T 1121.5-2006; exchangeable sodium (exchange. Na), potassium (exchange. K), magnesium (exchange. Mg), and calcium (exchange. Ca) were detected by referring to Chinese National Standard NY/T 1615-2008; exchangeable sodium percentage (ESP) denotes the percentage of exchangeable sodium content in cation exchange capacity, also known as alkalinity (Richards, 1954); TN was measured according to Chinese National Standard LY/T 1228-2015; total organic carbon (TOC) was measured according to Chinese National Standard HJ 695-2014; the detection of soil available potassium (Avail. K), available phosphorus (Avail. P), and available magnesium (Avail. Mg) were detected by referring to Chinese National Standard NY/T 889-2004/3.1, NY/T 1121.7-2014, and LY/T 2445-2015, respectively; other available metal ions were extracted by AB-DTPA method and determined by inductively coupled plasma atomic emission spectrometer (Prodigy Plus, Leeman, USA) (Soltanpour, 1985; Malathi et al., 2018). The soil around roots collected by ring-cutter was used for the following determinations: moisture content (MC), soil bulk density (SBD) in accordance with Chinese National Standard NY/T 1121.4-2006; the specific gravity was measured with reference to Chinese National Standard NY/T 1121.23-2010; porosity was calculated thus;

The total genomic DNA of organisms in soil samples were extracted by the CTAB method (Minas et al., 2011), and then the 16S rRNA gene V3-V4 hypervariable region in bacterial genomic DNA was amplified using V3-V4 universal primers. The sequence of V3-V4 universal primers were 341F-5’-CCTAYGGGRBGCASCAG-3’; 806R-5’-GGACTACNNGGGTATCTAAT-3’. DNA library construction was conducted using the DNA library construction kit (TruSeq^® DNA PCR-Free Sample Preparation Kit). After the library was qualified, high-throughput sequencing was conducted on the Illumina NovaSeq6000 platform of Beijing Novogene Technology Co., Ltd, China.

The physico-chemical properties of soil samples were statistically analyzed in SPSS 22.0 (IBM, US) software and further descriptive, variance, and Spearman correlation analysis were performed. The data between multiple groups were statistically tested by Duncan’s significance test (P< 0.05). Linear fitting of the data was undertaken using Origin 9.0 software (OriginLab, USA).

To achieve the readability of high-throughput sequencing data, the sequencing data processing was presented as follows: according to the Barcode sequence and the PCR amplification primer sequence, the sample data were separated from the off-line data; with the barcode and primer sequences removed, Flash V1.2.7 software (Magoč and Salzberg, 2011) was used to splice the reads sequences of each sample, and the spliced sequence then contained the raw rags; these were strictly filtered using Qiime V1.9.1 (Bokulich et al., 2013) to obtain high-quality tags (clean tags); clean tags removed chimeric sequences using Vsearch 1.2.5 software (Haas et al., 2011; Rognes et al., 2016), to obtain the final effective tags. Operational taxonomic unit (OTU) clustering and species annotation of effective tags: Uparse V.0.1001 software was used to cluster effective tags with 97% identity into OTUs (Edgar, 2013); species annotation analysis of OTU sequences was performed using the SSUrRNA database (http://www.arb-silva.de/) (Wang et al., 2007).

The visualization analysis of high-throughput sequencing data was described as follows: the indices of bacterial alpha diversity (e.g. Observed-species, Chao1, Shannon, Simpson) were calculated by using Qiime software; FAPROTAX software was used to predict bacterial function. Various packages in R 3.2.3 software were used for different analyses as follows: Venn diagrams were plotted using the Venn package; principal component analysis (PCA) diagrams and histograms of bacterial community composition were plotted using the ggplot package; significance testing of difference (T-test) among bacterial communities of different samples was analyzed using the rstatix, reshape2, and dplyr packages; redundancy analysis was undertaken using the vegan package; heatmap analysis was performed using the Hmisc and pheatmap packages. The WGCNA package in software R and Gephi 0.9.2 software were jointly used to study the bacterial co-occurrence network, in which the points in the network are bacteria at the genus level, and the network did not contain genera with the number of sequenced sequences less than, or equal to, 10 in different replicates; robustness calculation of the co-occurrence network was performed by R 3.2.3 software (Yuan et al., 2021). SPSS 22.0 software and SPSS Amos 24.0 software (IBM, US) were jointly employed to establish the requisite structural equation model (SEM).

The effects of different soil exogenous treatments on soil environment were significant, and indirectly affected the growth of S. cannabina. The analysis of soil mineral composition after different improvement treatments ( Figure S1 ) showed that the addition of organic fertilizer and river-sand did not change the main mineral composition of the soil, which was still dominated by quartz, muscovite, and albite; the additional serpentine is chrysotile, and the chemical formula is Mg₃[Si_{2- x} O₅](OH)_{4-4 x} . Firstly, the chemical characteristics of soil in different treatment sample plots were investigated as shown in Figures 1A , B . No matter whether S. cannabina grows, the order of soil saline-related environmental factors (total salt, EC) content in different treatment sample plots is such that CK > OF > FS > SS, and the order of soil alkalinity in different treatment sample plots is CK > FS > OF > SS ( Figure 1A ). In the bare soil without S. cannabina growth, the order of fertility contents in different treatment sample plots is OF > FS ≈ SS > CK; however, affected by the growth of S. cannabina, the order of fertility contents in different treatment sample plots is FS > SS ≈ OF > CK ( Figure 1B ). Secondly, the physical properties of the soil around the S. cannabina root ( Figure 1C ) are as follows: compared with the sample of CK, the sample OF had significantly increased soil porosity and moisture content (MC), but the soil bulk density (SBD) decreased; compared with OF, the sample of FS had significantly reduced soil MC; compared with the group of FS, the sample of SS had significantly decreased soil porosity and MC, and increased SBD. In addition, the order of biomass of S. cannabina growing in different treatment sample plots is such that SS > FS > OF > CK ( Figures 1C, D ).

Coastal areas are affected by saline intrusion, of which saline-alkali is the main stressor affecting agricultural production and ecology of coastal soil. Therefore, the specific components of salt-alkali-related cations in the soil around S. cannabina root under different improvement treatments were detected ( Table 2 ). The results showed that the contents of available cations and exchangeable cations in SS were significantly different from those in other treatments as follows: the contents of available base ions (Na, K) and trace elements (Fe, Mn, Cu) decreased significantly, while the contents of available Mg, Ca, and Ni increased significantly; the total amount of cation exchange in SS was significantly reduced, accompanied by significant decreases of exchangeable Na, K, and Ca, while the content of exchangeable Mg was remarkably increased.

Coastal saline-alkali silt soils are generally subjected to poor permeability and saline-alkali stress (An et al., 2022). According to the second national soil survey and relevant standards (National Soil Survey Office, 1998), the original soil on the experimental plot was subjected to severe salt stress (total salt > 6 g·kg^-1), while the soil nutrients are at the sixth level of extreme deficiency (TN< 0.5 g·kg^-1, TOC< 6 g·kg^-1) ( Figure 1 ). Therefore, with single application of organic fertilizer in CSS, the soil fertility and the biomass of S. cannabina were significantly improved ( Figures 1B –D ), improving the CSS.

Many studies (Fliessbach et al., 2007; Wang et al., 2016; Li et al., 2020) have shown that, short-term or long-term application of organic fertilizer in soil will lead to the improvement of soil microbial diversity. However, in this study, the Sesbania rhizosphere bacterial community of S. cannabina in the original CSS had a greater abundance ( Figure 2A ), more complex co-occurrence networks, and higher modularity ( Figure 5A ; Table 3 ). These results indicate that the single application of organic fertilizer causes sudden changes in soil physico-chemical characteristics ( Figure 2 ), which may strengthen the adaptive screening based on specific functional bacterial species in the original soil microbial library. The transformation of bacterial community structure in the rhizosphere of S. cannabina was affected by the single application of organic fertilizer with the type of nutrient utilization as the screening model. Among them, some widely reported organic matter degrading bacteria were significantly enriched ( Figures 3D, E ), such as Salinimicrobium and Gramella (Yun et al., 2018; Voisin et al., 2020); meanwhile, Bacteroidia and Actinobacteria (Gurushankara et al., 2019; Voisin et al., 2020) became the hub node of the bacterial co-occurrence network ( Figures 5B ; Tables S2–S5 ), and the mutual association in the bacterial co-occurrence network increased; in addition, the functional proportion of chemoheterotrophic and phototrophic in rhizosphere bacterial community of S. cannabina was significantly increased ( Figure 4 ; Table S1 ). The decomposition and utilization of soil organic matter forms the basis for soil bacteria to exert more biogeochemical cycling functions (Condron et al., 2010). In this study, the treatment with single application of organic fertilizer made more bacteria of organic matter decomposition and utilization screened and enriched, while these bacteria also played an important role in connecting the center of the bacterial network, which is conducive to soil organic matter activation and material circulation.

With single addition of organic fertilizer in CSS, the treatment can improve the soil fertility and significantly enrich functional bacteria that are conducive to the activation of soil organic matter; however, the moisture content has significantly increased ( Figure 1C ), and the soil salt stress has not been effectively mitigated ( Figure 1A ), which still needs further improvement.

Compared with the single application of organic fertilizer, in the treatment of organic fertilizer combined with river-sand (the simple case of FS), the moisture content and saline-alkali stress of soil were significantly decreased ( Figures 1A , C ); the growth of S. cannabina was further improved ( Figures 1C, D ). Meanwhile, functions of organic matter activation in the rhizosphere bacterial community of S. cannabina in FS were further enhanced ( Figure 4 ; Table S1 ), including: functions of ureolysis and aromatic_compound_degradation of the bacterial community were significantly enhanced ( Table S1 ); and then, the dominant bacteria of organic matter degradation had changed, such as Parasegetacter, Porphyromonas, o_Micrococcales, and f_unidentified_ Peptostreptococcales_ Tissirellales (Yan et al., 2020) and other bacteria were significantly enriched ( Figures 3A, B , D, E ). However, the treatment of FS increased functions of human pathogens and animal parasites _or_ symbionts, while the dominant bacteria such as Campylobacter, Pantoea, and the biomarker Massilia were reported as human pathogens (De Baere et al., 2004; Park and Shin, 2013). Hutchison et al. (2004; 2005) found that the application of organic fertilizer is one of the routes to contamination of soil by human pathogenic bacteria. In this study, the soil environment under the improved model of organic fertilizer combined river-sand may provide more suitable living conditions for these bacteria, which should draw attention to the relevant biological safety. Therefore, the quality management of organic fertilization regimes should be strengthened to prevent the introduction of pathogenic bacteria in large quantities with the application of organic fertilizer.

The changes of soil properties such as soil salinity, fertility, and physical properties are mainly displayed among samples of different treatments ( Figures 1 A–C ; Table 2 ). The application of organic fertilizer combined with river-sand and serpentine, as the sample of SS, highlights the crystal peak of serpentine in soil mineral crystals ( Figure S1 ). According to the results of the mineral chemical composition analysis (based on X-ray fluorescence) of serpentine in the serpentine mining area of Donghai County, Jiangsu Province (Lu et al., 2022), the main components are MgO (37.84%), SiO₂ (41.71%), Fe₂O₃ (14.37%), and CaO (3.29%). The weathering of minerals in SS may be the main reason for the significant increase of Avail. Mg and Exchange Mg in soil ( Table 2 ). In addition, compared with FS, the moisture content and saline-alkali stress (Aval. Na, Aval. K, and Exchange Ca) of soil in SS were significantly reduced ( Figures 1A, B , Table 2 ), and the biomass of S. cannabina was further significantly increased ( Figures 1C , D ). Simple linear model of RDA ( Figure 2D ) showed that the bacterial community composition in SS was correlated with Avail. Mg, moisture content, and Avail. Na. The covariance matrix of variables was used to study the correlation and multiple dependencies between multivariable (Lee, 2007; Larsson et al., 2020). Based on this, this study further estimated the causal relationship and multivariate correlation among the changes of environmental factors, the growth status of S. cannabina, and the rhizosphere bacterial community of S. cannabina, which are subjected to the influence of SS treatment, thus a structural equation model (SEM) between different factors was constructed, as shown in Figure 6 .

The fitting data of SEM shown in Figure 6 are good, as χ ² ≥ 0, df ≥ 2, 0.05< P< 0.90, and RMSEA< 0.05, indicating that the model is reliable. SEM demonstrated that the application of organic fertilizer combined with river-sand and serpentine significantly improved permeability and salinity of soil and was directly and strongly correlated with the construction of rhizosphere bacterial community of S. cannabina. At the same time, the biomass of S. cannabina was promoted through direct and indirect effects associated with improving the salinity. Among them, the rhizosphere bacterial community of SS was significantly different from that of FS, which could be explained in the following four ways:

In addition, Ni, at a low concentration, is the main component of [NiFe]-hydrogenase, which is used to recover H₂, a by-product of N₂ reduction in the nitrogen fixation process of legume rhizobia (Evans et al., 1988); meanwhile, Ni can mediate the up-regulation of ammonia transporter, the main metabolite of nitrogen metabolism (Liu et al., 2021), thereby promoting nitrogen cycle. In this study, the function of nitrate reduction in SS was significantly enhanced ( Figure 4 , Table S1 ), and the abundance of the rhizobium member Ensifer (Gopalakrishnan et al., 2015; Yan et al., 2016) that could attach to the root surface was significantly increased ( Figures 3A , C ). Ensifer has also been shown to increase photosynthesis and growth activity of plants (Peng et al., 2002; Chi et al., 2005; Toyama et al., 2021).

In summary, the positive development and functional analysis of the bacterial community in the rhizosphere of S. cannabina in SS are illustrated in Figure 7 . The diagram highlights the soil bacterial action caused by the addition of serpentine, based on the application of organic fertilizer and river-sand.

Affected by the changes in soil physico-chemical properties, especially the weathering of serpentine ( Figure 7 I ), many nitrate-reducing bacteria, nitrogen-fixing bacteria ( Figure 7 II ), and phototrophic bacteria ( Figure 7 III ) with promoted effects of plant growth were screened and enriched in the rhizosphere of S. cannabina in SS. Microorganisms can regulate bioavailable nitrogen by functions of nitrogen fixation and denitrification to exchange nitrogen with the atmosphere (Nelson et al., 2016), in which the enriched bacteria in SS can mobilize these two functions, which is conducive to improving the effect of nitrogen fixation in SS. In addition, the robustness of the co-occurrence network in SS was enhanced ( Figure 7 IV ), which can better withstand the interference of external or internal uncertainties. Although no increase in inorganic carbon was detected in the sample of SS, but the positive development in rhizosphere bacterial community of S. cannabina was conducive to the improvement of CSS and growth of S. cannabina, which was of significance for improving the effect of carbon sink in soil.