Soil microbial biomass
The total soil microbial biomass (SMB) of 86 soil samples from 30 study sites (Table S1, Fig. S1) across southern Europe Mediterranean regions was estimated using total soil dsDNA as a proxy (Table S2) 24 25. The amount of extracted metagenomic dsDNA from soils of the study sites ranged from 7.09 to 208.6 µg g-1 soil dry weight (Table S2) and varied significantly across sites (Fig.S2a. Kruskall-Wallis χ² = 11.42, p = 0.003), with soils of Extremadura (SP) presenting significantly higher SMB than soils from Sicily (IT) and Alentejo (PT) (Dunn test p < 0.05; Fig. S2). SMB was positively related to SOC and also to TN (Table 1) coherently with the positive correlation between SOC and TN (Plassart et al., 2019; Grilli et al, 2021). SMB was higher in non-arid sites (Fig. S3, soils with AI > 0.65, Wilcoxon X² = 11.61, p <0.001), as a result of higher plant productivity, increased access to substrates due to soil moisture, and higher microbial growth rates, in accordance with the results found in other studies 8 26. SMB also varied across cover with coniferous stand soils displaying the highest SMB, and croplands the lowest (Dunn test p < 0.05, (Fig. 1, Table S2). Microbes in croplands are generally affected by cultivation practices that disrupt soil aggregates, where they live, and by inorganic fertilization, that increases nutrient availability, promoting microbial community shifts towards more copiotrophic taxa, with higher respiration rates; 27. Since most SOC formation is mediated by microbial activity 28, a reduction in microbial biomass and, eventually, in its substrate use efficiency, in tilled soils, may diminish the overall potential for SOC formation 29. For these reasons sustainable soil management, based on reduced tillage, organic fertilization and cover crops, could help to maintain crop yield while avoiding further SOC depletion in semi-arid soils30.
Interestingly high SMB was detected under the coniferous forest of the driest investigated area in Lampedusa (IT1, aridity index: 0.29), where high SOC levels were also detected 10 in agreement with other studies on SOC under pine tree woods 31. The levels of microbial biomass under coniferous stands is controversial because other studies report that soils under pine plantations, in semi-arid soils, present markedly smaller and less active microbial communities 32 33 34, generally explained by the lower availability and degradability of organic substrates, such as lignin 24. However the involvement of bacteria in wood litter decomposition including cellulose and aromatic compounds derived from lignin degradation is more important than was previously assumed 35. As bacteria, more than fungi, can modulate at community level their carbon use efficiency, the specific composition of the soil microbiota could explain SOC accumulation in dry soils. Drought could act on bacteria by reducing the microbial respiration rate 36 and increasing the stability of soil organic matter 15. Moreover bacteria could persist in soils for long periods in a dormancy condition as seed banks and be revived when environmental conditions change 37 7. All this could contribute to higher resilience to drought of coniferous forest ecosystems in respect to other soil uses 38 39. We found the lowest SMB levels in soils with SOC < 20g kg -1 (Wilcoxon Z = -4.73, p < 0.001), the threshold value below which any additional decline of SOC concentrations might result in a steep variation of most soil parameters 10, indicating that this concentration is also critical for key microbe-driven soil ecosystem services 9.
Relationship between bacterial diversity and edaphic variables
Molecular fingerprinting of the soil bacterial communities, was obtained by the automated ribosomal intergenic spacer analysis (ARISA) on a subset of 69 soil DNA samples from the 30 study sites. Average OTU richness ranged from 9.0 (S.E. 0.5) to 46.5 (0.26) and Shannon index ranged from 2.34 (0.26) to 3.55 (0.2) (Table S2). Soil bacterial richness and Shannon index significantly differed across the sites (Fig. S2, ANOVA F = 33.08, p < 0.001, Kruskall-Wallis χ² = 20,68, p < 0.001) with soils of Sicily presenting higher richness than all the other sites (p < 0.001 after Tukey HSD). Richness and Shannon index also varied across land cover (F = 4.79, p = 0.002; χ² = 21,76, p < 0.001), although only for croplands and shrublands values were significantly different (after Tukey HSD; Fig. 1). Bacterial richness was higher in coniferous soils while Shannon index was higher in croplands; both indexes showed the lowest values in shrubland soils.
Among the analyzed soil variables, pH was the most influencing factor affecting soil bacterial alpha-diversity with significantly positive relationships (p < 0.001), conversely it had no significant effect on SMB, which was positively related with SOC (Table 1).
Bacterial diversity tended to be higher in the low-SOC soils (SOC<20g kg-1), where Shannon diversity reached the highest values (Z = 2.28, p = 0.02), while bacterial richness was not significantly influenced by low SOC levels (t = -1.72, p = 0.08). Coherently, semi-arid sites (AI< 0.65) presented higher bacterial diversity (richness t = -5,04, p < 0.0001; Shannon index X² = 11.78, p <0.001) than non-arid sites (AI>0.65) (Fig. S4). Several studies have reported such a negative relation between soil bacterial diversity and total C, leading to the conclusions that, a decline in SOC, as a result of soil degradation and aridity, leads to an increased bacterial diversity 1 40 41 especially in arid soils 9. In the semi-arid soils, however, a large part of the biomass, contributing to the extracted dsDNA, could be dormant, thus metagenomics could overestimate the real expressed soil functional microbial diversity. In this respect metaproteomics could be a better predictive indicator of ecosystem functionality than DNA diversity 37.
Soil TN was inversely related to both bacterial OTU richness and Shannon index (Table 1). TN is not always considered an explanatory variable for microbial diversity 22 although our data suggest that it can have a stronger negative relationship with bacterial diversity than SOC.
The Canonical Analysis of Principal coordinates (CAP) indicated that beta-diversity is related to the main soil parameters 10 (Fig. 2), mainly pH, SOC and CEC, as already reported in other studies 1 41.
The similarities among the bacterial communities across all sites reflected their biogeography confirming that soil bacterial communities were spatially structured 35 22 22. Interestingly, the coniferous forest soils (SP2-1 and IT1-2) displayed the highest distance from their respective biogeographic cluster, indicating the strong influence on the soil bacterial assemblage of this soil cover, in respect to the others.
Table 1. Summary of the regression testing the effects of SOC, pH, N and CEC on soil microbial biomass (SMB), richness (estimated as observed bacterial OTUs, n=86), and diversity (Shannon index) calculated on ARISA profiles and measured as soil extracted dsDNA (n=69). R² = adjusted fit of the model. In bold characters
|
|
Estimate
|
SE
|
t
|
p
|
SMB
|
Intercept
|
-19.156
|
28.005
|
-0.68
|
0.496
|
R² = 44
|
SOC
|
0.572
|
0.177
|
3.23
|
0.002
|
|
pH
|
7.013
|
4.089
|
1.72
|
0.091
|
|
TN
|
10.012
|
3.974
|
2.52
|
0.014
|
|
CEC
|
-0.889
|
0.798
|
-1.11
|
0.270
|
Richness
|
Intercept
|
-3.368
|
6.723
|
-0.50
|
0.618
|
R² = 0.49
|
SOC
|
0.099
|
0.042
|
2.33
|
0.023
|
|
pH
|
4.676
|
0.982
|
4.76
|
< 0.001
|
|
TN
|
-2.249
|
0.954
|
-2.36
|
0.021
|
|
CEC
|
0.334
|
0.192
|
1.74
|
0.086
|
Shannon
|
Intercept
|
2.498
|
0.240
|
10.393
|
< 0.001
|
R² = 0.21
|
SOC
|
0.002
|
0.001
|
1.648
|
0.105
|
|
pH
|
0.091
|
0.035
|
2.612
|
0.011
|
|
TN
|
-0.068
|
0.034
|
-2.022
|
0.048
|
|
CEC
|
0.005
|
0.007
|
0.712
|
0.480
|
The soil microbiota of European Mediterranean soils
Illumina sequencing of the 16S rRNA gene was carried out on a subset of 22 soil DNA samples. The subset was selected to include different soil covers in the three regions, and two replicates (a and b) for each site. Only one sample (IT1-2 replicate b) gave no results, due to low quality DNA that hampered further processing. The subset of 21 soil DNA samples was sequenced and resulted (after denoising and chimeras check) in a minimum of 13,482 to a maximum of 73,358 reads, with an average of 44,922 reads for each sample (Table S3). Richness, assessed on bacterial observed OTUs clustered at 97%, ranged from 107 to 581 OTU97 (Table S3), both extremes were observed in two croplands in central Sicily (Italy). OTU97 richness, was positively correlated to the alpha diversity based on ARISA profiles (Pearson tests r = 0.36, p = 0.12) considering all data, and the correlation improved (r = 0.56, p = 0.013) excluding the outlier IT4-1a. This significant correlation shows that the two data sets of bacterial diversity were congruent 42.
All soils contained the main typical soil bacterial phyla40 but with different abundances in respect to European soils of other climatic zones 22 and also in respect to decertified soils of other continents 41.
The most abundant bacterial phyla were Proteobacteria (average relative abundance 26.6%), Actinobacteria (24.2%), Acidobacteria (16.6), Firmicutes (11.3), Bacteroidetes (5.5), Chloroflexi (5.4) and Verrucomicrobia (5.0) (Fig. 3). The dominance level of these phyla is more similar to previous characterizations of world’s dry forests and drylands40 than non-Mediterranean European soils 22. In Southern European soils Proteobacteria were less abundant, and not always the dominant phylum, while Actinobacteria and Acidobacteria were much more abundant than in higher European latitudes 22. Archaea were below 2%, mainly represented by Crenarchaeota. Beyond Proteobacteria, the phyla Acidobacteria, Verrucomicrobia and Chloroflexi were more abundant in the soils of Extremadura (SP) and Alentejo (PT) while Actinobacteria were more abundant in Sicilian soils (IT) (Fig. S3). The overall relative abundance of the 7 main phyla did not vary across soils (Kruskal-Wallis X² = 0.76, p = 0.68) or land use (X² = 1.09, p = 0.89). We did not find any significant difference in the relative abundance of each phylum among the five different land cover (ANOVA and Kruskal-Wallis p > 0.05). Considering each phylum separately, however, the abundance of 5 out of the 7 dominant phyla was significantly related to cation exchange capacity (CEC), with positive and negative relationships (Table 2). Soil CEC is considered strongly correlated with the organic carbon content 22 as well as to the physical (structural stability), chemical (nutrient availability), and biological characteristics of soils 43. Four phyla were related to pH, two positively, Firmicutes and Verrucomicrobia and two negatively, Acidobacteria and Chloroflexi. Interestingly, the phylum Firmicutes was significantly related to all edaphic variables (SOC, pH, TN, and CEC), conversely Proteobacteria was not significantly related to any of the analyzed variables. In particular the relative abundance of Firmicutes was negatively related to SOC, in agreement with the role of Firmicutes as responsible for enhanced mineralization of dissolved organic carbon and increased soil respiration 44 45.
Table 2. Summary of the OLS and GLM models testing the effects of cation exchange capacity (CEC), pH, total nitrogen (TN) and soil organic carbon (SOC) on the relative abundances of the 7 most abundant phyla. In bold characters the significant values (p < 0.05).
Phylum
|
|
Estimate
|
SE
|
t / z
|
p
|
Acidobacteria
|
Intercept
|
26.78
|
4.25
|
6.30
|
< 001
|
|
CEC
|
-0.28
|
0.11
|
-2.49
|
0.024
|
|
pH
|
-1.60
|
0.68
|
-2.35
|
0.032
|
|
TN
|
0.42
|
0.42
|
0.98
|
0.340
|
|
SOC
|
0.04
|
0.02
|
1.76
|
0.098
|
Actinobacteria
|
Intercept
|
-5.66
|
9.41
|
-0.60
|
0.556
|
|
CEC
|
0.67
|
0.25
|
2.68
|
0.016
|
|
pH
|
2.97
|
1.51
|
1.97
|
0.066
|
|
TN
|
-1.23
|
0.94
|
-1.31
|
0.210
|
|
SOC
|
0.05
|
0.04
|
1.10
|
0.287
|
Bacteroidetes
|
Intercept
|
1.57
|
0.99
|
1.58
|
0.114
|
|
CEC
|
0.05
|
0.03
|
1.97
|
0.049
|
|
pH
|
-0.04
|
0.16
|
-0.23
|
0.821
|
|
TN
|
-0.10
|
0.10
|
-1.01
|
0.313
|
|
SOC
|
0.00
|
0.00
|
-0.46
|
0.648
|
Chloroflexi
|
Intercept
|
6.46
|
0.92
|
6.98
|
< 001
|
|
CEC
|
0.07
|
0.03
|
2.69
|
0.007
|
|
pH
|
-0.70
|
0.16
|
-4.39
|
< 001
|
|
TN
|
-0.11
|
0.09
|
-1.30
|
0.194
|
|
SOC
|
-0.01
|
0.00
|
-2.21
|
0.027
|
Firmicutes
|
Intercept
|
0.55
|
1.06
|
0.52
|
0.602
|
|
CEC
|
-0.11
|
0.03
|
-3.95
|
< 001
|
|
pH
|
0.39
|
0.16
|
2.37
|
0.018
|
|
TN
|
0.39
|
0.12
|
3.19
|
0.001
|
|
SOC
|
-0.02
|
0.01
|
-2.37
|
0.018
|
Proteobacteria
|
Intercept
|
17.09
|
7.16
|
2.39
|
0.030
|
|
CEC
|
0.16
|
0.19
|
0.84
|
0.413
|
|
pH
|
0.41
|
1.15
|
0.36
|
0.727
|
|
TN
|
-0.02
|
0.72
|
-0.03
|
0.979
|
|
SOC
|
0.05
|
0.03
|
1.57
|
0.136
|
Verrucomicrobia
|
Intercept
|
17.37
|
3.40
|
5.11
|
< 001
|
|
CEC
|
0.02
|
0.09
|
0.21
|
0.836
|
|
pH
|
-1.87
|
0.54
|
-3.43
|
0.003
|
|
TN
|
0.20
|
0.34
|
0.60
|
0.557
|
|
SOC
|
-0.03
|
0.02
|
-1.76
|
0.098
|
Conversely, Acidobacteria and Chloroflexi, more abundant in acidic soils of Spain and Portugal, under broad leaved agroforest, have low carbon turnover and, consequently, low carbon dioxide emission and higher carbon sequestration capacity 46. In soils of Sicily the most abundant phylum was Actinobacteria 7. The definition of this phylum as copiotroph or oligotroph is controversial. Some studies associate it with copiotrophy, others suggest that it is better suited to oligotrophic conditions probably because it includes members that are able to use both labile and recalcitrant substrates 47. In areas with SOC below 20g kg-1 Actinobacteria (t = 2.70, p = 0.014), Bacteroidetes (Z = -2.10, p = 0.035) and Proteobacteria (t = 2.91, p = 0.012) were significantly less abundant suggesting a copiotrophic lifestyle of these three phyla, including Actinobacteria.
The correlation of edaphic variables with soil bacterial diversity and lifestyle, however, depends on the taxonomic level considered 47. In fact, we observed that some critical parameters, as CEC, can have no correlation with phyla as a result of positive and negative relations with taxa at the order level within the same phylum (Table S4). In general, the correlation between the chemico-physical parameters and the five most abundant orders within each phylum, confirmed the prevalent role of pH in modulating the distribution of bacterial taxa in soils, with positive and negative correlations (Table S4). Within Actinobacteria two orders (Acidimicrobiales and Solirubrobacterales) showed positive and significant relation to SOC, but, in contrast, the sole genus Streptomyces highly abundant in low SOC soils showed negative correlation to SOC and also to CEC. Within Proteobacteria, Rhodospirillales and Rhizobiales were instead positively related to SOC suggesting that they may contribute to C sequestration in soils of semi-arid regions. Only 38% of the sequences (on average) was identified at the genus level (Fig. 4).
Most of the genera with abundancies above 1% were correlated to soil pH with the exclusion of those within the phylum Firmicutes (Table 3). Conversely, in the Firmicutes, 4 out of the 6 most abundant genera (Bacillus, Ammoniphilus, Solibacillus and Sporosarcina) were negatively correlated to SOC and 3 also to CEC, with Ammoniphylus and Bacillus showing highly significant negative correlations (Table 3). The negative correlation with SOC of genera in the Firmicutes, was also confirmed at the order level (Table S4). Bacillus and Ammoniphilus, are known copiotrophs meaning that they rapidly consume the substrate 48 thus giving low contribution to SOC accrual. The genus Bacillus in particular, was dominant in low-SOC pastures, broad-leaved forest soils of Portugal and in a cropland of central Sicily, where Opuntia ficus indica is organically cultivated with no-tillage and mulching using Opuntia cladodes (M. Russo personal communication). The abundancy of copiotrophs in this cropland could be the result of abundant fresh organic matter (but with low molecular diversity) resulting from mulching of Opuntia cladodes. Probably introducing a greater molecular diversity by mulching with more than one organic source, may increase the metabolic demand, and thus potentially limit soil C loss by respiration 49. However, before considering Firmicutes as markers of low quality soils, we should also consider the role of Bacillus as a plant growth promoter and P solubilizing agent, especially valuable in arid soils 50. Only two genera were positively correlated with CEC, namely Rubrobacter (Actinobacteria) and Candidatus Entotheonella (Deltaproteobacteria) both detected in italian soils and both also positively related to soil pH.
Table 3. Spearman correlation coefficients between the relative abundance of the most abundant genera and main soil parameters: cation exchange capacity (CEC), pH, total nitrogen (TN) and soil organic carbon (SOC) obtained in the sampled soils. Only significant correlations (p-value < 0.05) are shown. Blue and red colors indicate negative and positive relationships, respectively, with color intensity representing the degree of significance (p < 0.05 > 0.01, < 0.01 > 0.001 and < 0.001).
The core microbiota of semiarid Mediterranean soils
The soil core microbiota was identified as phyla and genera that exist in all the samples with at least 1% of the relative abundance within each sample, respectively 41. At the phylum level the core microbiota of semi-arid soils under desertification risk of southern Europe (considering exclusively those from sites with an AI<0.65), was composed of four phyla Proteobacteria, Actinobacteria, Acidobacteria and Firmicutes (Fig. 3). At genus level we were unable to define a common core microbiota, due to high variability of edaphic parameters. At regional level Bacillus, Rubrobacter and Balneimonas were the (identified) core microbiota in Sicilian soils (AI 0.29-0.49) (Fig. 4). While Bacillus is an ubiquitarious genus, Rubrobacter, belonging to a deep evolutionary line of descent in the class of Actinobacteria is known for its multi-extremophilic growth conditions. Although it was not possible to classify it at species level, this genus is highly represented in extremely hot, arid and/or acidic ecosystems or habitats with severe radiation/desiccation conditions, such as deserts, volcanic areas and other arid regions 30 51. Its presence was recently associated to high polyphenol oxidase activities and accumulation of polyhydroxyalkanoates that could function as protection against stress factors 38. Balneimonas (Bradyrhizobiaceae) is also associated to arid soils where it produces extracellular material that contributes to the formation of soil crusts 52; 53.
The core microbiota of Alentejo soils (AI 0.35-0.38) beyond Bacillus, included: DA101, Candidatus Solibacter (Acidobacteria), and Rhodoplanes (Fig.4). The phylotype DA101 within the, as yet, underdescribed soil phylum Verrucomicrobia, has been reported in a wide range of soils and ecosystem types throughout the world. In this work it resulted negatively correlated to CEC, coherently with its previous identification as low fertility specialist 54. Similarly, Candidatus Solibacter belonging to slow growing oligotrophs adapted to resource limitations 35, is strongly negatively correlated to soil pH and CEC (Table 2 and 3). This genus reported as dominant in arid soils 53, was abundant also in coniferous forest soils of our study sites. Rhodoplanes is a phototrophic 53, nitrogen fixing genus 55.
These traits, together with the positive correlation with SOC, suggest a pivotal role of this genus in low nutrient Mediterranean semi-arid soils where it may be involved in carbon and nitrogen fixation. Rhodoplanes was also the only genus represented above 0.8% in all investigated soils. Thus, Rhodoplanes could be considered “the rare endemic genus” within arid Mediterranean soils where drought reduces the dominance of common bacterial species and allow rare soil bacteria suited for resource limited environments to proliferate 56 57.