Sodium SeO₃^2–, sodium hydroxide (NaOH), dipotassium phosphate (K₂HPO₄), monopotassium phosphate (KH₂PO₄), potassium persulfate (K₂S₂O₈), sodium sulfite (Na₂SO₃), n-caprylic alcohol, acetic acid, and hydrochloric acid were obtained from Sinopharm (Shanghai, China). SeNPs were synthesized and purified using a recently described method (Shakibaie et al., 2010). In brief, the SeNPs were biologically reduced from SeO₃^2– by microorganisms. The reduction product was washed and disrupted in liquid nitrogen. The suspension was then liquid–liquid extracted by n-caprylic alcohol. The purified SeNPs were collected from water phase. More details of synthesis purification and SEM observation were shown in Supplementary Text 1.

The soil used in this study was collected from a farmland in Jimo, Shandong province, and the physical and chemical properties of soil were shown in Supplementary Table 1. Five experiments with different Se treatment groups were established: (1) Control group, in which ddH₂O was applied; (2) Low SeO₃^2– group, in which 1 mmol/kg soil SeO₃^2– solution was applied; (3) High SeO₃^2– group, in which 10 mmol/kg soil SeO₃^2– solution was applied; (4) Low SeNPs group, in which 1 mmol/kg soil SeNP suspension was applied; (5) High SeNPs group, in which 10 mmol/kg soil SeNP suspension was applied. The experiments were conducted in a greenhouse at 22°C, and the photoperiod was 16 h light and 8 h dark. Three soil samples were collected before all treatments were applied and designated as samples for Day 0. Soil samples of each treatment were collected in triplicate on Days 7 and 30 for further analysis.

The contents and existing forms of Se in soil samples were determined using a modified continuous extraction method (Martens and Suarez, 1997; Wright et al., 2003; Kulp and Pratt, 2004; Keskinen et al., 2009). Briefly, 1,000-g soil sample was placed in a 50-ml centrifuge tube and extracted using different solutions with a 1:10 solid/liquid ratio for each step. The water-soluble Se was extracted using H₂O, and the ligand-exchangeable Se was extracted using 0.1 mol/L K₂HPO₄-KH₂PO₄ (pH 7.0). The alkali-soluble Se was extracted by 0.1 mol/L NaOH, and the elemental Se was extracted with the combination of 1 mol/L Na₂SO₃ (pH was adjusted to 7.0 by hydrochloric acid) and ultrasonication. The acid-soluble Se was extracted using acetic acid. The Se concentration in the extracts was detected with an atomic fluorescence photometer (AFS-933, Titan Instrument, Beijing, China). More details were described in Supplementary Text 2.

Here, 0.25-g fresh soil sample was weighed to extract the total soil DNA. The extraction was performed using a TIANamp Soil DNA Kit (TIANGEN Biotech, Beijing, China). The DNA extracts were quantified and quality checked using a NanoDrop One UV–VIS Spectrophotometer (Thermo Fisher Scientific, United States). The DNA extracts were stored at −80°C for further analysis.

The abundance of total bacteria was assessed by quantification of 16S rRNA gene copies through qPCR. A universal primer set K90q/K94q (Supplementary Table 2) targeting bacterial 16S rRNA was employed. The protocol of qPCR followed a previously described method (Liu et al., 2020).

The V4–V5 region of the bacteria 16S rRNA gene was amplified with the primer set 515F/907R (details shown in Supplementary Table 2). In this step, an 8-bp-long barcode unique to each sample was appended. The PCR product was purified and quantified for the library generation. Then, the amplicon library was sequenced using an Illumina MiSeq platform at Shanghai Biozeron Biological Technology Co., Ltd. (Shanghai, China). Raw sequencing data were uploaded in NCBI Sequence Read Archive (SRA) database under accession number PRJNA664643.

High-throughput sequences were processed using the QIIME2 platform (version 2019.10) (Bolyen et al., 2019). Briefly, raw sequencing data were demultiplexed, filtered, and denoised. The denoised sequences were then clustered into operational taxonomic units (OTUs) with a 97% similarity. The representative sequences of each OTU were aligned against the SILVA database (SSU 135) with a confidence threshold of 70% to identify the taxonomy (Quast et al., 2013).

Alpha diversity indexes including Chao1, Shannon, and Simpson indices were calculated following the Vegan package in R. Principal coordinates analysis (PCoA) was calculated following the ape and GUniFrac package in R. LEfSe and correlation network analysis was conducted through the MicrobiomeAnalyst platform (Chong et al., 2020). LEfSe analysis was calculated using a significance level of p < 0.05 and an LDA score of 2. Correlation network analysis was performed at the genus level using the SparCC method with a significance level of p < 0.05 and a correlation threshold of 0.8 (Friedman and Alm, 2012; Jiang et al., 2017). According to the species classification data, the effects of Se on the function of soil microorganisms were predicted using the Tax4Fun2 algorithm.

Fresh soil samples of 0.25 g were collected for EPR analysis. The samples were loaded into a quartz EPR tube, placed into the EPR resonator, and analyzed with a Bruker EMXnano spectrometer (Karlsruhe, Germany). The parameters were set as follows: center field, 3,420 G; sweep width, 80 G; sweep time, 40.06 s; receiver gain, 40 dB; modulation amplitude, 2.000 G; microwave frequency, 9.8 GHz; and microwave attenuation, 20.00 dB.

Two typical soil microorganisms, namely, Gram-negative bacteria Bacillus sp. and Gram-positive bacteria E. coli, were chosen to investigate responses of different soil bacteria to exogenous Se. Briefly, the two bacteria were inoculated in a Luria-Bertani (LB) medium at a ratio of 1:100. SeNPs or SeO₃^2– at a concentration of 0, 0.1, 0.5, 1, 5, and 10 mmol/L were added, respectively, and the mixed media were incubated at 37°C, 180 rpm for 24 h. OD₆₀₀ was measured to characterize the biomass. Reactive oxygen species level was determined using Reactive Oxygen Species Assay Kit (Solarbio, Shanghai, China) following the manufacturer’s instructions. Glutathione peroxidase (GSH-Px) activity was determined using a Micro Glutathione Peroxidase Assay Kit according to the manufacturer’s protocol (Solarbio, Shanghai, China). Protein concentrations were determined following the bicinchoninic acid assay method (Smith et al., 1985).

The transformation of SeNPs and SeO₃^2– in soil was analyzed with a continuous extraction method. As shown in Figure 1, the element Se decreased significantly along time in the High SeNP group from 248.58 mg/kg on Day 7 to 229.93 mg/kg on Day 30 (p < 0.001). In contrast, there was no significant difference in Se content in the Low SeNP group (29.98 mg/kg on Day 7, 28.99 mg/kg on Day 30). No soluble Se was detected in treatments with SeNPs. In both High SeO₃^2– group and Low SeO₃^2– group, the dominant form of Se was water-soluble SeO₃^2–, and its concentration decreased significantly along time (p < 0.001). Alkali-soluble Se (-II) was the second predominant Se form, which was reported to be mainly presented as organic-associated Se in soil (Keskinen et al., 2009). In addition, a relatively high elemental Se was observed in the High SeO₃^2– group. Studies have shown that SeO₃^2– is affected by redox potential in soil and can be reduced gradually by soil microorganisms, where the toxicity of Se decreases with decreasing valence (Piacenza et al., 2017; Wang et al., 2017; Zhai et al., 2021).

To explore the effect of SeNPs and SeO₃^2– on the soil microbial community, soil samples were collected on Days 7 and 30. The abundance of soil bacteria was quantified by qPCR. The abundance of bacteria in the Control group remained relatively stable at 2.78–2.37 ^∗ 10¹¹ copies/g soil (Figure 2A). In the Low SeNP group, the abundance of soil bacteria increased by 146% (Day 7) and 171% (Day 30) compared with that of the Control group. A relatively smaller increase in the number of bacteria was observed in the High SeNP group, which was 129 and 136% on Days 7 and 30, respectively. However, the abundance of soil bacteria significantly decreased in treatments with SeO₃^2–, which was decreased by 45% (Day 7) and 55% (Day 30) in the Low SeO₃^2– group. In the High SeO₃^2– group, the higher concentration of SeO₃^2– exhibited a stronger adverse effect on soil bacteria, where the abundance of bacteria was decreased by 70% on Day 30 compared with that of the Control group.

In this study, SeNPs promoted the development of soil microorganisms. It has been reported that SeNPs can play an antioxidant role in organisms by enhancing the activities of key enzymes in the antioxidant defense system (Zhai et al., 2017), activating antioxidant selenase (Guo et al., 2020), inhibiting protein glycosylation (Yu et al., 2015), or directly scavenging free radicals in vivo, therein promoting microbial growth. SeNPs are more stable in the environment and exert lower toxicity to microorganisms as a biological reduction product. However, the transformation of SeO₃^2– in and out of microorganisms usually produces reactive oxygen species (Kessi and Hanselmann, 2004) that cause the toxicity of SeO₃^2–.

The effects of SeNPs and SeO₃^2– on soil microbial community structure were further analyzed by 16S rRNA amplicon sequencing. A total of 1,488,614 sequences were obtained and assigned to 4,069 OTUs according to 97% similarity. No significant difference in α diversity was observed between the SeNP groups and the Control group (Figure 2B and Supplementary Table 3). According to PCoA of β diversity distance between samples (Figures 2C,D), the distance between the High/Low SeNP groups and the Control group was smaller than that between the SeO₃^2– groups and the Control group. These results indicated that SeNPs exhibited no significant disturbance to the diversity of the soil microbial community.

However, there was a significant difference in the α diversity indexes between the High/Low SeO₃^2– groups and the other groups (p < 0.01). The observed OTU number, Chao1 index, Simpson index, and Shannon index were all obviously reduced (Figure 2B and Supplementary Table 3). The High/Low SeO₃^2– groups were separated from the other groups (Figures 2C,D). These results indicated that SeO₃^2– affected the species and abundance of soil microorganisms. Further analysis of β diversity revealed a significant separation between the SeO₃^2– groups and the Control group in the PCoA results based on the Bray–Curtis distance, which correlates with species abundance (Bray and Curtis, 1957; Figure 2C). However, PCoA results based on unweighted UniFrac distance, which is related to the developmental composition of species (Lozupone et al., 2011), showed no significant grouping differences, and the degree of interpretation among the principal components was low (Figure 2D). The different results of the two diversity distances implied that the SeO₃^2– caused an abundance decrease but not extinction of soil microorganisms. It suggests that the effect of Se on soil microflora may not aim at specific species but be broad spectrum to all the soil microorganisms.

As shown in Figure 3A, Proteobacteria dominated in soil samples from all treatment groups, with a relative abundance exceeding 50%. Significant differences were observed between the experimental groups at the genus level (Figure 3B). To quantitatively analyze the species changes, LEfSe method was used to analyze the species abundance at the genus level on Days 7 and 30 (Figures 3C,D), aiming to screen the species with significant differences in each experimental group. Under the given conditions, 205 and 267 significantly different species were identified, respectively. The relative abundance of several genera, e.g., Acidibacter, Acidipila, and Bradyrhizobium, was higher in the Control group, indicating that their relative abundance decreased under the stress of SeNPs and SeO₃^2–. The relative abundance of Rhodanobacter, Bacillus, Alicyclobacillus, and Granulicella increased in the SeNP groups, while the relative abundance of Chujaibacter, Candidimonas, Jatrophihabitans, and Nevskia increased in the SeO₃^2– groups. The increased relative abundance may be due to the promotion effect of SeNPs or Se resistance of microorganisms.

The influence of SeNPs and SeO₃^2– on the soil microbial community was studied with a co-occurrence network method (Figure 4). Based on SparCC method, the correlation of soil microbial abundance was calculated, and then the graph inference of the network and the estimation of several topological characteristics were carried out. In the SeNP groups, the association network has 89 nodes and 308 connections, which can be classified into 16 constituent units (Figures 4A,C). In the SeO₃^2– groups, there were 119 nodes, 442 connections, and 25 basic units (Figures 4B,D). To summarize, the impact of SeNPs on soil microbial network is less than that of SeO₃^2–. Based on intermediate centrality, the key microorganisms in each network were identified. In SeNP groups, Tuberibacillus and Telmatospirillum were the microorganisms located in the core of the network, while Bryobacter and Mizugakibacter were the microorganisms located in the core of the network in SeO₃^2– groups.

The enriched microorganisms (e.g., Bacillus, Alicyclobacillus, Granulicella) in SeNP groups have been previously reported as soil probiotics (Zelenikhin et al., 2013; Long et al., 2018). In addition, Tuberibacillus and Telmatospirillum at the core of the microbial network are involved in the decomposition of soil organic matter and utilization of carbon, nitrogen, phosphorus, and other elements (Hatayama et al., 2006; Xue et al., 2017; Ding et al., 2019), indicating that the addition of SeNPs induces active probiotics related to basic soil functions and improves the soil microbial community. Chujaibacter, Candidimonas, Jatrophihabitans, Nevskia, Bryobacter, and Mizugakiibacter, which were enriched in the SeO₃^2– groups or lived in the core of the network, have the ability to treat heavy metals and resist UV, antibiotics, and other biological or abiotic stresses (Vaz-Moreira et al., 2011; Větrovský and Baldrian, 2015; Prakash et al., 2020; Sánchez-Soto et al., 2020). These results imply that SeO₃^2– appeared to be toxic to the soil microbial community.

The effects of exogenous Se on soil functions were predicted based on the Tax4Fun2 algorithm (Supplementary Figure 2). There was no significant difference between treatments, indicating that exogenous Se exhibited little impact on soil function. Therefore, the soil microbial community can effectively resist the external interface and maintain the ecological functions of the soil, reflecting the robustness of the soil microbial community function.

In this study, exogenous SeNPs and SeO₃^2– could change soil microbial community obviously. To explore the influencing mechanism of SeNPs and SeO₃^2– on the soil microbial community, persistent free radicals in the soil were determined, as the migration and transformation of Se may induce the release of free radicals (Kessi and Hanselmann, 2004). By analyzing the G-factor of free radical signal, it was found that the persistent free radicals in soil were mainly oxygen-center free radicals and semi-quinone free radicals (Khachatryan and Dellinger, 2011; Figure 5 and Supplementary Table 4). There was no significant difference in free radical signals between the SeNP groups and the Control group. However, in the SeO₃^2– groups, the persistent free radical signal in the soil was significantly enhanced on Day 7, but there was no significant difference on Day 30. Nevertheless, the bacterial community showed a more significant difference on Day 30, indicating that the corresponding response of the soil microbial community to exogenous Se compounds lagged behind the migration and transformation of exogenous Se in soil. The migration and transformation of exogenous Se in the environment, especially SeO₃^2–, is mainly due to chemical reactions, but the changes of soil microbial community need a longer period compared with chemical reactions.

In this study, the application of SeNPs and SeO₃^2– exhibited different effects on the soil microbial community. SeNPs promoted the propagation of soil probiotics, while SeO₃^2– induced the enrichment of microorganisms with a higher tolerance to environmental stress. In addition, we found that the relative abundance of Gram-negative bacteria increased more than that of Gram-positive bacteria after SeNP and SeO₃^2– treatments. Therefore, we selected two representative soil microorganisms (Bacillus sp. and E. coli) to investigate their response to exogenous Se. As shown in Figure 6A, the microbial biomass increased with the increase of SeNP concentration, while SeO₃^2– at different concentrations had little influence on the microbial biomass. Regardless of the Se concentration, both tested bacteria showed higher growth in treatments with SeNPs compared with that of SeO₃^2– groups. In addition, experimental groups with E. coli always showed higher biomass than that with Bacillus sp. Similar to the results of biomass, the total protein concentration in SeNP treatment groups was higher than that of SeO₃^2– treatment group, and the total protein content of E. coli was higher than that of Bacillus sp. (Figure 6B). The above results indicated that SeNPs promoted the growth of microorganisms at a certain concentration compared with SeO₃^2–, and the promotion was more significant in E. coli. We further determined the intracellular reactive oxygen species of different treatment groups and found that the intracellular reactive oxygen species of Bacillus sp. were lower than that of E. coli (Figure 6D). It implies that E. coli had a higher ability to resist oxidative stress, which is consistent with its higher biomass.

GSH-Px is a kind of antioxidant enzyme with Se atom as its active center, and its activity can characterize the bioavailability of Se (Yu and Wei, 2013). The GSH-Px activity in the SeNP groups was significantly lower than that in the SeO₃^2– treatment groups (Figure 6C). Although GSH-Px activity was significantly different (p < 0.05), there was no significant correlation between enzyme activity and Se concentration (Figure 6C and Supplementary Figure 3). The GSH-Px activity of Bacillus sp. was higher than that of E. coli, partly resulting in the lower amount of intracellular reactive oxygen species in Bacillus sp.

The results indicated that SeNPs were less bioavailable than SeO₃^2–. It has been reported that SeO₃^2– can be easily transported across the membrane via binding to the sulfhydryl sites on the cell membrane or by phosphate transfer protein (Yu et al., 2018; Zhu et al., 2020). In contrast, SeNPs are in a reduced state. The Se atom is coated with a layer of extracellular polymer (Avendaño et al., 2016), making it difficult to enter the microorganism directly. It may partly explain the lower toxicity of SeNPs.