Soil-intake or sterilized-soil-intake increased gut microbial diversity and richness
The four groups of mice were treated for 42 days as shown in Fig. 1. Fresh fecal samples from mice were collected (Fig. S1) and the gut microbiota analyzed using high-throughput sequencing.
We sequenced the 16S rDNA of the fecal samples with polymerase chain reaction (PCR) amplification of the V4 hypervariable region primed using the 515F-907R primer set. For every quarter increase in microbial diversity or richness, the risk of allergic disease is reduced by 55% [30]. A total of 1,921,486 sequences were qualified with the average sequence of each sample being 50,565. We used rarefaction to normalize the number of reads in each sample of the operational taxonomic unit (OTU) table to 38,680 sequences and analyzed the diversity of the intestinal microbiota in each group. The rarefaction curve revealed that the microbiota diversity of the Soil group and SS group were similar, which were higher than those of the MW group and Control group, which were similar (Fig. 2A).
As shown in Fig. 2B and Table S4, the number of OTUs in the SS group was significantly more than that in the MW or Control groups (P < 0.05), but less than that in the Soil group (P < 0.05). There was no significant difference between the MW and Control groups. Similar results were obtained when abundance was estimated using the Chao1 index (Table S4 and Fig. 2C). However, the Shannon index showed no significant difference between the SS and Soil groups (Table S4 and Fig. 2D).
From this viewpoint, the intake of sterilized soil significantly improved the diversity and richness of intestinal microorganisms in mice to levels similar to that of the Soil group. However, neither the diversity nor richness of the gut microbiota significantly changed when only microbes isolated from soil were ingested.
Changes in gut microbial composition
Principal coordinates analysis (PCoA) of the unweighted UniFrac distance matrix showed obvious differences between each group of mice. Specifically, the intestinal microbial structure of the mice in the SS and Soil groups were similar, whereas those in the MW and Control groups were similar (Fig. 3A). It could also be seen that the distance between the SS and Soil groups was shortest, followed by that between the Control and MW group (Fig. 3B, Table S5). The longest distance was between the MW and Soil groups. Furthermore, PCoA of the Bray-Curtis matrix of metagenomics sequences were like that of the 16S rRNA gene sequences (Fig. 3C and 3D, and Table S5).
The most abundant genera based on 16S rDNA sequences were identified using the RDP Classifier. There were obvious differences in the composition of the top 20 most abundant genera among the four groups (Fig. 3E and Table S6). The pie chart of the Control group was more similar to that of the MW group, whereas that of the SS group was more similar to that of the Soil group. The same conclusions were reached using the column of genera abundance/type in each phylum of the top five most abundant phyla (Fig. S2) and the results of species abundance/type among the different groups (Fig. S3).
Random forest is a supervised machine learning technique that uses multiple decision trees to train and predict samples. It is a powerful classifier that can use non-linear relationships and complex dependence between OTUs/strains to identify the OTUs/strains that are important to the structural makeup of the microbiota. An importance score is assigned to each OTU/strain based on the increased error caused by deleting that OTU/strain from the prediction set. In the current study, we considered an importance score for an OTU of at least 0.0005 as being highly predictive. There were 67 predictive OTUs at the species-level between the SS and SPF groups, of which 55 (82%) were overrepresented in the SS group (Table S7). However, there were 36 predictive OTUs between the MW and SPF groups with only 25 OTUs (69%) being overrepresented in the MW group (Table S7). Correspondingly, analysis of the Soil and SPF groups showed there were 74 predictive OTUs with the Soil group presenting 63 (85%) more OTUs (Table S7). Interestingly, compared with MW group, there were 64 and 49 predictive OTUs in Soil and SS groups, respectively, among which 54 (84%) were overrepresented in Soil group and 41 overrepresented in the SS group (Table S7).
The results of random forest analysis of strains based on shotgun sequencing of the microbial metagenome were similar to those of the above 16S rDNA sequencing analysis (Fig. S4 and Table S8). The strains listed in Table S8 meet two standards; first, the random forest importance score of the strain was at least 0.001, and second, the p-value of t-test was less than 0.05. Compared with Control mice, there were 128 predictive species for the Soil mice, 108 for the SS mice, and 52 for the MW mice, overrepresenting 96, 85, and 47 microbes, respectively (Fig. S4A and Table S8).
Based on these results, we concluded that ingestion of soil-isolated microbes, sterilized soil, or farm soil each had an influence on the intestinal microbial structure and composition of mice. The effect of eating sterilized soil was more similar to that of the Soil group, whereas the effect of drinking soil microbes was more similar to that of the Control group.
Intake of sterilized soil increased the abundance of type III secretion system (T3SS) genes
To further understand the mechanism by which ingesting soil or drinking soil microbes influenced the mouse intestinal microbiota, analyses of microbe species and genes were conducted using the metagenomic shotgun sequence data. To search for their biological relevance, we selected microbe species with significant differences between experimental groups based on t-test analysis and with importance scores exceeding 0.001 (Table S8). It was determined that the biological functions for 62.7–89.8% of the microbes selected had not been previously published (Fig. S4B, Table S8). Among the species selected, four had functions reported. One is an engineering bacterium, the second is an antibiotic producing microbe, the third is pathogenic to plants or animals other than mice, and the fourth one is a mouse pathogen (Fig. S4B, Table S8). Compared to that of the Control mice, the SS and Soil mice intestinal microbiota were more abundant for mouse pathogens (Fig. 4A, Fig. S4B, and Table S8), whereas the MW mice had no greater number of mouse pathogens but had more of four other types of microbes (Fig. S4B and Table S8).
When the abundance of functional genes was compared between groups, it was found that the SS group showed a significant increase in the abundance of genes encoding T3SS and two-component systems compared with that in the Control or MW mice (Fig. 4B, Fig. S5 and Table S9). Compared to the MW mice, the Soil mice harbored more genes encoding T3SS (Fig. S5), two-component systems, and butanoate metabolism (Fig. 4B and Table S9). Accordingly, the ingestion of soil or sterilized soil increased the abundance of mouse and other pathogens, as well as genes coding for T3SS and two-component systems.
Intake of soil microbes increased the abundance of genes for short-chain fatty acid metabolism and amino acid biosynthesis
Compared with that of the Control mice, the MW mice exhibited increased abundance of enzyme genes used in the metabolism of short-chain fatty acid (Fig. 5A and Table S9), including butyric acid, propionic acid, and acetic acid, as well as genes involved in amino acid biosynthesis. Compared with that of the SS mice, the Soil mice exhibited similar differences (Fig. 5B and Table S9).
The Soil mice ingested the same soil as the SS mice, but the soil-based microbes remained for the Soil group. Compared with that of the Control group, the Soil group of mice demonstrated increased abundance of not only genes for T3SS and two-component systems but also that for short-chain fatty acid metabolism and amino acid synthesis (Fig. S6). Furthermore, the abundance of genes for flagellar assembly (Fig. S7) and bacterial chemotaxis (Fig. S8), as well as additional genes for short-chain fatty acid metabolism (Fig. S9) and amino acid biosynthesis, was also increased in the Soil mice (Fig. S6).
The intake of soil microbes played an important role in increasing the abundance of genes involved in short-chain fatty acid metabolism and amino acid biosynthesis. Besides the common functions induced by the intake of either soil microbes or sterile soil, the intake of soil containing the microbes prompted the germination a greater number of different functional genes in addition to their increased relative abundance.
Soil-intake or sterilized soil-intake decreased serum IgE levels
To analyze the impact of ingesting soil, sterilized soil, and soil microbe-containing water on the immune function of mice, we stimulated eczema on the skin of the four experimental groups of mice using DNFB and then measured serum IgE levels. The results revealed the serum IgE levels of the Soil and SS mice were significantly lower than those of the Control mice (P < 0.05). Furthermore, the levels of the Soil mice were significantly lower than the MW mice (P < 0.05; Fig. 6A). Although the median IgE value of the SS group was lower than the MW group, the difference was not statistically significant (Fig. 6A). Skin damage was also scored for the mice. The skin lesion scores in the Soil and SS groups were significantly lower than those in the Control and MW groups (P < 0.05), but there was no significant difference between the Soil and SS mice or between the MW and Control mice (Fig. 6B).
The Soil group and SS group demonstrated significant increases in the numbers of mouse pathogens and genes of T3SS. To determine whether these pathogens led to infection, hematological analysis of blood samples was performed. The results showed no indication of infection (Table S10).
To further explore which strain or gene was related to the enhancement of immune function in then mice, we performed a correlation analysis between IgE levels and the strains or functional genes detected by high-throughput sequencing (Fig. S10, Fig. S11, Table S11, and Table S12). The results showed that the microbes with significant correlation included fritillary virus Y and the soil bacteria Burkholderia glumae, among others (Fig. S10 and Table S11). The significance of these microbes in mice has not been reported, nor have there been any reports on their influence on immunity. The significant correlation for functional genes included six genes of T3SS and six genes of metabolic pathways (Fig. S11 and Table S12). The T3SS genes came from the same cell organ and work together to perform their function in promoting infection by the bacteria. Therefore, these T3SS-coding genes may be more closely related to the levels of IgE than the genes for the metabolic pathways.
There were 76 genes of bacterial secretion systems registered in the Kyoto Encyclopedia of Genes and Genomes (KEGG) database of which 43 (56.6%) were sequenced in this study. As shown in Table S13, 15 (34.9%) of the genes in the Soil group and 18 (41.9%) of the genes in the SS group were significantly more abundant than those in the Control group (P < 0.05). Only three genes of the MW group were observed at abundance higher than the Control group. Compared with that of the MW group, 11 (25.6%) genes of the Soil group and 9 (20.9%) genes of the SS group were more abundant.
Overall, the results showed that the intake of soil or sterilized soil improved the immune function of mice and did not cause obvious infections. The immune function was positively and correlated with the bacterial secretion system genes, especially with that of T3SS.