Results

\label{results}

Soil physico-chemical properties

\label{soil-physico-chemical-properties}

Site effect on soil physico-chemical-biological properties

\label{site-effect-on-soil-physico-chemical-biological-properties}
The site appeared as the main factor affecting the soil physico-chemical parameters (Table 1, a). The PCO allows to visualize the direction of change in soil parameters between the both sites (Figure 1). The site effect on each parameter is detailed in the main test of Table 2. The texture was as expected, i.e. the sandy site was characterized by a higher sand content, while the silty site was characterized by a higher silt and clay content. In terms of nutrient status, the silty site showed a higher fertility level with a higher P and Ca content as well as a higher CEC, rate of base saturation and pH, whereas the sandy site showed a higher C:N ratio. In terms of physical properties, the quantity of macroaggregate was higher in silty site, whereas the stability of aggregates and the silt+clay fraction was higher in sandy site. Finally, the fungal richness was strongly impacted by the site, showing a higher value in sandy site.

Farming effect on soil physico-chemical properties

\label{farming-effect-on-soil-physico-chemical-properties}
The farming effect on soil physico-chemical-biological parameters was lower when compared to the site effect (Table 1, a). In addition, a marginally significant interaction effect between site and farming was noticed (Table 1, a), suggesting that the farming effect slightly relies on the type of soil. Based on that results, we analyze the farming effect on each site independently. The site-based PCO allows to visualize the direction of change in soil parameters between the both farming systems (Figure 2). The overall farming effect as well as the site-related farming effect on each parameters is detailed in the main test and pairwise test of Table 2, respectively. We also noticed that farming-sensitive soil parameters were not the same according to the site.
The sandy soil was characterized by a higher humidity, HWC and total biomass content under organic farming and a higher phosphorus, and silt+clay content under conventional farming. A slight increase of microaggregates quantity and a better aggregate stability was also noticed under organic farming (Figure 2, Sandy site). The silty soil was characterized by a higher potassium, magnesium, silt and clay content under conventional farming, and a higher aggregate stability under organic farming (Figure 2, Silty site). Although the discrimination between conventional and organic farming seems to be stronger in silty site, the pairwise test of Table 1 revealed a stronger farming effect in sandy soil. These results may be explained by the samples corresponding to “CONV-3” in sandy site , where their distance from the rest of samples was high (Figure XX, Sandy site), and mainly due to the particularly dried soil conditions.

Spatial component effect on soil physico-chemical properties

\label{spatial-component-effect-on-soil-physico-chemical-properties}
The spatial component (plot) ranked as the second main factor affecting the soil physico-chemical properties (Table 1, a). A strong interaction effect between the farming system and the spatial component was noticed, suggesting a strong influence of the spatial component on the farming effect. On PCO, the distance between each farming system of each pair (represented by the form of the symbol) varies.

Soil microbial community diversity

\label{soil-microbial-community-diversity}
We obtained 1 070 378 bacterial 16S (29 733±3780 per sample) and 2 404 498 fungal ITS2 (66 792±13 815 per sample) high-quality sequences from the 36 samples. Sequence clustering yielded 7 643 (3227±268 per sample) bacterial and 2 681 (751±86 per sample) fungal OTUs.

Overall site effect on microbial diversity

\label{overall-site-effect-on-microbial-diversity}
The site appeared as the main driver of beta diversity for both bacteria and fungi (Table XX, a). The site was able to explain 34% and 20% of bacterial and fungal community variability, respectively (Table XX, a). PCOs allow to visualize the major site effect on community composition, represented by the discrimination of samples on the first axis (Figure XX). At high taxonomic level, the relative abundance of some dominant taxa such as Alphaproteobacteria, Actinobacteria, Verrucomicrobia and Zygomycota was higher at the sandy site, while some other dominant taxa such as Chloroflexi, Deltaproteobacteria, Gammaproteobacteria, Acidobacteria, Betaproteobacteria, Planctomycetes and Glomeromycota showed a higher abundance at the silty soil (Figure XX).
In terms of alpha diversity, fungi were more rich and less even at the sandy site as compared to the silty site (Figure XX and Table XX, a). The site effect on bacterial alpha diversity, however, was not significant (Table XX, a).

Overall farming effect on microbial community diversity

\label{overall-farming-effect-on-microbial-community-diversity}
The farming effect ranked as second driver of beta-diversity for both bacteria and fungi (Table XX, a). The farming system was able to explain 11% and 18% of bacterial and fungal community variability, respectively (Table XX, a). PCOs allow to visualize the farming effect on community composition, represented by the discrimination of samples on the second axis (Figure XX). For bacteria, the difference in community composition was only due to the dissimilarity instead of heterogeneity of variance. For fungi, however, the difference observed in community composition may also be attributed to variance heterogeneity (supplementary data, Permdisp). Consequently, the PCO helps to visualize the partitioning of these two sources of difference. Here, the discrimination between the farming groups (organic and conventional) can be easily visualized on PCO (Figure 2) leading to conclude that the difference observed in fungal community composition was likely due to the studied factors. As the farming effect on community composition was detected at high taxonomic, we identified the farming-sensitive taxa. Some dominant taxa, such as Bacteroidetes , Chloroflexi, Actinobacteria, Alphaproteobacteria , Gammaproteobacteria and Ascomycota were showed to be more abundant under conventional farming. Some others, however, such as Nitrospirae, Acidobacteria , Planctomycetes, Deltaproteobacteria, Basidiomycota and Glomeromycota, were more abundant under organic farming (Figure XX). While they are less abundant, some bacterial taxa recently recognized as Candidate phylum such as TM7, WS2, OP11, OP3 and WS3 showed a significant change in their relative abundance between organic and conventional farming. The putative ecological roles provided by these farming-sensitive taxa to the agroecosystem will be discussed below with respect to the current literature.
In terms of alpha-diversity, the farming system had less impact than for the beta-diversity. The only significant farming effect was detected on fungal richness: overall, organic farming was more rich and less even than the conventional farming (Figure XX).

Farming effect interacts with site and spatial component

\label{farming-effect-interacts-with-site-and-spatial-component}
The magnitude of farming effect was influenced by the site (Table XX, interaction terms between farming and site). The pairwise test (Table 1, b) showed a stronger farming effect (lower avg sim) on beta-diversity at the silty site when compared to the sandy site, for both bacteria and fungi. PCO allows to visualize the stronger farming effect at the silty site, as the distance between SiltCONV and SiltORGA was higher than the distance between SandCONV and the SandORGA (Figure XX).
In the same way, the magnitude of farming effect was influenced by the spatial location (Table XX, interaction terms between farming and plot). The pairwise test for each pair of plot is provided in Table XX, b. Consequently, the farming effect on beta-diversity should be analysis with respect to the site and spatial location.
In the same way for alpha-diversity, the farming effect on bacteria tended to be stronger in silty site than in sandy site (Figure XX). For fungi, however, the magnitude of farming effect tended to be similar in the both sites, as no significant interaction between farming and site was noticed (Table XX, Figure XX). As soon as the farming effect was significant, the organic farming showed a higher value of richness, but a lower value of evenness. An interactive effect with the spatial location on bacteria and fungi was also detected. The respective pairwise test for each pair of plot is provided in Table XX, b.

Relationship between community composition and soil physico-chemical-biological properties

\label{relationship-between-community-composition-and-soil-physico-chemical-biological-properties}
Some physico-chemical-biological properties were closely related to the microbial beta-diversity variability. The best descriptors of bacterial community composition included the silt content, the pH, the silt-clay fraction, the HWC, the macroaggregates, the aggregate stability, the microaggregate, the sand and clay content, the bacterial richness and evenness, the N and K content, and the fungal richness. In the same way, the best descriptors of fungal community composition included the silt content, the aggregate stability, the microaggregates, pH, the silt-clay fraction, the macroaggregates, the HWC, the sand and clay content, the bacterial richness and evenness, the fungal richness, the total biomass and the nitrogen content.
Among the chemical soil parameters, pH, HWC, N and K were identified to be good descriptors of the bacterial beta diversity variability. For fungi, the analysis identified
The distLM analysis identified the particle size (clay, silt and sand), pH, soil humidity, and some nutrients (nitrogen, phosphorus, calcium and magnesium) to be the best predictors of the bacterial and fungal community composition (Figure XX). The model successfully explains the main trends of the bacterial and fungal variance displayed in the constrained ordination (Figure XX).
As depicted by the fungal dbRDA, particles size (clay, silt and sand), humidity, pH, calcium and phosphorus clearly discriminate the fungal communities between the silt and the sand site. The discrimination between the organic and conventional farming, however, is not obvious: magnesium seems to be the best variable to separate conventional from organic farming. Regarding the bacterial dbRDA, there was also a clear discrimination between the silt and sand site mainly drove by the particle size (clay, silt and sand) and humidity. Here, pH, calcium and phosphorus were good predictors of organic farming at the silt site, while magnesium seems here again to be the best to separate conventional from organic farming.
Beta-diversity
\label{beta-diversity}
The site appeared as the main driver of beta diversity, explaining 34% and 20% of bacterial and fungal community variability, the farming system ranked second, explaining 11% and 18% of bacterial and fungal variability, and the spatial component (plot) ranked third, explaining 11% and 13% of bacterial and fungal variability, respectively (Figure 2 and Table 1, a). The PCO allows to visualise the difference in community composition between the sites and the farming systems (Figure 2). The first axis separates the silty and sandy site, while the second axis separates the conventional and organic farming system. For bacteria, the difference in community composition was only due to the dissimilarity instead of heterogeneity of variance. For fungi, however, the difference observed in community composition may also be attributed to variance heterogeneity (supplementary data, Permdisp). Consequently, the PCO helps to visualize the partitioning of these two sources of difference. Here, the discrimination between the site groups (silty and sandy), the farming groups (organic and conventional) and spatial component (plots) can be easily visualized on PCO (Figure 2) leading to conclude that the difference observed in fungal community composition was likely due to the studied factors.
A strong influence of site on the farming effect was evidenced, as reported by the interaction effect between site and farming (Table 1, a). The pairwise test (Table 1, b) showed a lower average of similitude (avg sim), i.e. a stronger farming effect on beta-diversity at the sandy site when compared to the silty site. Moreover, the contrasted farming effect on community composition between the both site was stronger for fungi. PCO allows to visualize the stronger farming effect at the silty site, as the distance between SiltCONV and SiltORGA was higher than the distance between SandCONV and the SandORGA (Figure XX).
The spatial location had also a strong influence on the farming effect, as reported by the interaction effect between plot and farming (Table 1, a). The detail of the farming effect for each pair of plot is provided in the pairwise test (Table 1, b).
Alpha-diversity
\label{alpha-diversity}
The sandy site tend to be more rich and less even as compared to the silty site for both bacteria and fungi (Figure XX, Table XX, a). Fungi were more sensitive to the site and farming system effects, whereas bacteria was more sensitive to the spatial location (plot) (Table XX, a). As reported by the significant interaction term (Table XX, a), the farming effect on bacterial alpha-diversity tended to be stronger in silty site than in sandy site (Figure XX). For fungi, however, the farming effect tended to be similar in the both sites, as no significant interaction between farming and site was noticed (Table XX, Figure XX). As soon as the farming effect was significant, the organic farming showed a higher value of richness, but a lower value of evenness. ⇒ Permdisp analysis