Soils samples were collected from homefields and outfields in the village of Diohine, which is located in the Senegalese Groundnut Basin at 14˚29'51"N, 16˚30'36"W [23] . The local soil is classified as Arenosol [35] , with mostly low clay contents (<5%; mainly kaolinite) [24] . The climate is Sudano-Sahelian, characterized by a long dry season (October to June) and three months of rainy season from July to September. Annual rainfall is 530 mm, and the average annual temperature is 30˚C. Woody vegetation is dominated by Faidherbia albida distributed in parkland; the geological substratum consists of tertiary sandstones [24] . The homefields and outfields whose soils we sampled produced millet (Pennisetum typhoides): the area’s main crop, which in some cases is rotated with maize, cowpea, and groundnut.

Two types of plots were used (homefields and outfields) that were representative of the two cultural practices in groundnut basin. Homefields received substantial amounts of cattle and small ruminant manures every year or two ranging 4 to 20 t·ha⁻¹ of dry matter while outfields received organic fertilization ≤ 1 Mg ha⁻¹·yr⁻¹ [23] .

The soil samples were collected during the dry season at depth of 0 - 10 cm in six plots of homefields and six plots of outfields (Supplementary material 2). As shown in Supplementary material 3, the homefields soils contain 0.9% clay, 2.2% silt, 96.9% sand, 0.45% SOM, 0.26% total C, 1.88 meq/100g CEC, 0.02% total N, and 4.84 ppm assimilable P. Their C/N ratio is 10.9 and the pH 6.72. The outfields soils contain 1.3% clay, 1% silt, 97.7% sand, 0.22% SOM, 0.13% total C, 1.30 meq/100g CEC, 0.01% total N, and 3.75 ppm assimilable P, with a C/N ratio of 12.9 and pH 5.86. Soils’ physico-chemical characteristics were performed at ISRA, CNRA laboratory of Bambey, Senegal (https://www.isra.sn). For the characterization of SOM and C the modified Walkley and Black [36] , method was used. The modified Olsen method was used for assimilable P, and Kjeldahl method for total N. The Robinson’s pipette method with USDA classification was used for soil texture, the ammonium acetate for CEC, and a laboratory pH-meter with electrode and extractor for the pH.

After the physico-chemical analysis, the six replicates of each practice were, pooled, sieved to <2 mm and stored at room temperature pending processing. The sample for each site consisted of six pooled subsamples.

The experiment was carried out under glasshouse controlled conditions. In the region of interest in Senegal CP4-Africa simulations (AMMA-CATCH, 2018) (Supplementary material 1), showed that temperature can exceed 50˚C at surface of the soil. Therefore, we chose 60˚C as the extreme heat-stress temperature so that we could determine which of the microbial communities is most resistant to extreme thermal stress. Further support for choosing 60˚C is found in Riah-Anglet et al. [37] , who report not only that microbial communities’ activities are affected similarly by heat stress at 50˚C and 60˚C, but that 60˚C represents an extreme thermal stress [1] [37] [38] [39] .

The heat-stress experiments and subsequent analyses were done on aliquots of six homefields and six outfields soils. The aliquots for a given pooled soil sample (i.e., six pooled subsamples of homefields soil or outfields soil) were prepared as follows. First, we determined the pooled sample’s residual water content and water-holding capacity (WHC), after which we added demineralized water to raise the sample’s water content to 80% of WHC. The aliquots’ samples were then pre-incubated at 28˚C for seven days to stabilize their microbial activity, per the recommendations of Wada and Toyota [21] .

After pre-incubation, the microcosms were made of 330 ml glass bottles filled with 30 g of equivalent dry aliquot soils at 80% WHC [6] . Each bottle was then sealed with a waterproof rubber plug. Heat stress in sealed bottles does not resemble actual heat stress in the field, but several studies have used this method to elucidate the different responses of the microbial community to disturbance and thus give trends of potential change in the real environment [1] [6] [40] [41] . We simulated extreme heat stress under controlled conditions and applied two distinct treatments: control (without stress) and heat stress. As a control sample (i.e., not heat-stressed), and following the protocols of Wada and Toyota [21] , one of the bottles was then held at 28˚C while the others were held for (variously) 3, 14, or 28 days at 60˚C to simulate heat stress [1] [37] [38] . The three durations are denoted as SD₃, SD₁₄, and SD₂₈, respectively. We ran three replications of each duration for each pooled soil sample. After each heat-stress sample had completed its time at 60˚C, it was given a 28-day recovery incubation at 28˚C [42] .

Each bottle was weighed every 3 days, from the beginning of the heat stress until the end of the post-stress recovery incubation. Demineralized water was added as needed to maintain the soil moisture content between 70% and 80% of WHC. At the end of incubation, we took triplicate soil samples from each bottle for molecular analyses, then stored the 3 samples at -80˚C until DNA extraction.

To quantify the metabolic activities of soil microorganisms after the heat stress [43] , we measured basal respiration (i.e., the CO₂ emitted by soil samples) [14] [17] of the same aliquots of soil (i.e., those that were sealed in 330-ml bottles) that are described in Section 2.2. CO₂ levels in the bottles were measured using gas phase micro-chromatography (μ-CPG Agilent 490, 1109602). Air was renewed frequently via an air pump to avoid accumulation. The total amounts of CO₂ emitted (µg C-CO₂ g⁻¹ soil) were calculated after the 28-day recovery incubation for each of the 3 stress durations (SD₃, SD₁₄, and SD₂₈). For each of those durations, C mineralization rates (µg C-CO₂ g⁻¹ soil·day⁻¹) of a given soil sample were measured during the 28-day recovery incubation. The rate was also calculated immediately after heat stress.

We calculated Orwin and Wardle’s RS and RL indices, which are generic ones that may be used for any sort of disturbance [5] to estimate the microbial communities’ resistance and resilience under heat stress.

The RS index for a given soil aliquot and heat-stress duration was calculated from C mineralization rates that were measured immediately after heat stress. That is, at the end of SD₃, SD₁₄, and SD₂₈:

RS ( t 0 ) = 1 − 2 | D 0 | / ( C 0 + | D 0 | ) (1)

where D₀ is the difference between P₀ (in our case, the C mineralization rates of the heat-stressed sample immediately after heat stress) and C₀ (the rate for the control samples; i.e., those which were maintained at 28˚C while the others were undergoing heat stress). Note t₀ is the moment at which the rates were measured at the end of heat stress.

The resilience index, RL, was calculated from C mineralization rates that were measured at time t_x, the end of the 28-day recovery incubation that followed SD₃, SD₁₄, and SD₂₈:

RL ( t x ) = 2 | D 0 | / ( | D 0 | + | D x | ) − 1 (2)

Here, D₀ is as above, and D_x is the difference between the C mineralization rates of the control soil (C_x) and the heat-stressed soil (P_x) at the time point (t_x) chosen to measure resilience. Note that the values of these two indices are bounded by −1 and +1 with a value of +1 showing that the disturbance had no effect (maximal resistance), and lower values showing stronger effects (i.e. less resistance). An index value of 0 indicates either a 100% reduction or increase in the value of the disturbed soil. Similarly, an RL of +1 indicates that P_x = C_x (complete recovery after the disturbance), and lower values indicate slower recovery. An index value of 0 indicates that the disturbed soil has either not recovered at all since the disturbance ended (i.e. D₀ = D_x) [5] .

After 28 days of recovery incubation, high-throughput sequencing was performed on SD₂₈ of homefield and SD₁₄ of outfield to find the microbial communities responsible of the partial resilience observed at the end of their 28-day recovery. For the homefield, three heat stress samples of SD₂₈ and control without incubation and three heat stress samples of SD₁₄ and control without incubation of outfield were used for DNA extraction.

Thus, total genomic DNA of each soil sample was extracted from 0.25 g of soil using the FastDNA^TM SPIN kit for Soil (MP Biomedicals, CA, USA), with modification of the manufacturer’s instructions [44] .

The quality and concentration of the extracted DNA was verified after electrophoresis migration on 1.5% agarose gel. High-throughput sequencing was performed at ADNID (Montferrier, France; http://www.adnid.fr) with Illumina MiSeq system (Illumina) targeting 16S rRNA gene with the 515F/806R primers set and ITS gene with the ITS3F-ITS4R primers. The sequences were deionized, and operational taxonomic units (OTU) were defined by clustering at 3% divergence (97% similarity) followed by removal of singletons and chimeras. Final OTUs were taxonomically classified using BLASTn against a curated database derived from GreenGenes and SYLVA. We then produced the final OTUs tables containing the number of sequences per sample per OTU matching the designated taxonomic classification. The whole process was conducted at ADNID (Montferrier, France; http://www.adnid.fr)

The statistical analyses were performed with R software v. 3.1.3 (Peter Dalgaard, CET 2015). Normal distribution of residuals and homogeneity of variance were assessed by (respectively) the Shapiro and Bartlett tests. If these two conditions were met, one-way ANOVA was performed to analyze the effect of heat stress. A Tukey HSD test was used for pairwise multiple comparisons if heat-stressed samples and control samples differed significantly in their physico-chemical characteristics, C mineralization, or microbial-diversity indices. A Kruskal Wallis non-parametric test was performed whenever residuals were not normally distributed, or variances were inhomogeneous. Each sample’s α-diversity of bacterial and fungi communities was evaluated by calculating richness and the Shannon and Simpson indexes.

CO₂ accumulation (µg C-CO₂ g⁻¹ soil) was calculated after 28 days of recovery incubation. For all 3 heat-stress durations, C mineralization in homefield samples was significantly higher than that of the controls (by 32.19% for SD₃, 86.94% for SD₁₄, and 94.12% for SD₂₈) (Table 1). In the heat-stressed outfield samples, the C mineralization was again higher than in the controls (by 2.53% for SD₃; 18.3% for SD₁₄; and 8.96% for SD₂₈), but only the SD₁₄ sample’s increase was significant. C mineralization in the control homefield samples is also 2.26 to 2.5 fold the C mineralization in the control outfield samples, comparable to organic C ratio between homefield and outfield (2.69).

Heat-stressed homefield soils had high mineralization rates compared to control. The average rate for SD₃ was 8.83 µg C-CO₂ g⁻¹ soil·day⁻¹, versus 6.96 and 5.36

Superscripts indicate significant differences between stress and control samples (p < 0.05), n = 3.

for SD₁₄ and SD₂₈, respectively. All 3 of these rates were significantly higher (p < 0.05) than the low, stable rates of the corresponding control samples (2.22, 1.84, and 1.59 µg C-CO₂ g⁻¹ soil·day⁻¹ respectively for SD₃, SD₁₄, and SD₂₈) (Table 2). However, mineralization rates decreased towards the end of the heat stress and during the recovery incubation (Figures 1(a)-(c)). At the beginning of that recovery, the SD₃ samples showed a respiration pulse 115.4% higher (significant at p < 0.05) than that of control samples (Figure 1(a)). In contrast, the SD₁₄ and SD₂₈ treatments showed no such difference in respiration flux between control and heat-stress samples (Figure 1(b), Figure 1(c)).

In the case of outfield soils, the average mineralization rates of heat-stress samples were again higher than the low, stable rates of the controls (1.44, 0.85, and 0.81 C-CO₂ g⁻¹ soil·day⁻¹ for the heat-stressed SD₃, SD₁₄, and SD₂₈ samples, versus 1.22, 0.94 and 0.84 µg C-CO₂ g⁻¹ soil·day⁻¹ for the respective control samples). However, none of the differences are significant (Table 2). C mineralization rates decreased towards the end of the heat stresses and at during the recovery incubation (Figures 2(a)-(c)). In addition, the respiration pulses exhibited by heat-stress samples at the beginning of the recovery incubation were higher than those of the corresponding controls (157.9% higher for SD₃, versus 302% and 75.5% respectively for SD₁₄ and SD₂₈) (Figures 2(a)-(c)).

The carbon mineralization analyses were performed at the end of heat stresses (t0) and at the end of the 28-day recovery incubations (t28). The resulting mineralization rates were compared to those for control samples, via appropriate statistical analyses.

These indices, proposed by Orwin and Wardle [5] , are generic ones that can be calculated from different types of data to highlight moderate effects upon function that might not be revealed by simple statistical analyses. We calculated these indices from C-mineralization results, as described in Section 2.4.

Using >97% sequence identity, the high-quality reads were clustered into operational taxonomic units (OTUs). The bacterial and fungal α-diversities of homefield soils and outfield soils (as quantified by richness and the Shannon and Simpson indices) were quite different (Table 4).

In the homefield soils, bacterial OTU richness was significantly greater in the heat-stressed soils than in the control soils. In contrast, fungal OTU richness was significantly greater in the controls. In the outfield soils, OTU richness was significantly greater in the control soils for the bacteria as well as the fungi. These results indicate that in the homefield soils, heat stress increased the bacterial richness, but decreased the richness of fungi, whereas heat stress reduced the richness of both types of microbes in outfield soils.

Superscripts indicate significant differences between heat-stressed and control samples according to software R (3.1.3). (p < 0.05). n = 3.

The specific diversity, as quantified by the Shannon and Simpson diversity indices, varied across the heat-stress treatments. In the homefield soils, these indices and the OTU richness showed the same trends for the bacterial community. For fungi community in the homefield soils, the Shannon and Simpson indices for heat-stressed samples were not significantly different from those of control samples. However, in the outfield soils the values of the indices for fungi were higher in control samples than in the heat-stressed ones (Table 4).

In homefield and outfield soils alike, heat stress changed the relative abundances (as compared to control samples). The taxonomic inventory of the sequences identified 14 bacterial and 7 fungal phyla. The dominant bacterial phyla (relative abundance > 1%) were Firmicutes, Proteobacteria, Chloroflexi, Actinobacteria, Acidobacteria, Bacteroidetes, Gemmatimonadetes, Planctomycetes, and WPS-2 (Figure 3(a)). The major phyla of fungi were Ascomycota, Basidiomycota, Chytridiomycota, and Mucoromycota (Figure 3(b)).

At the class level, we found thirty bacteria classes, all of which were present in every sample, and belonged mostly to the phyla Firmicutes, Chloroflexi, Proteobacteria, Acidobacteria, and Actinobacteria. In the non-stressed samples from homefield, the bacterial and fungal classes Bacteroidia, Blastocatellia, Gemmatimonadetes, Nitrososphaeria, δ-Proteobacteria, Chytridiomycetes, and Sordariomycetes were significantly more dominant than in the non-stressed samples from outfield. In the latter samples, the dominant classes were Acidobacteriia, Bacilli, Ktedonobacteria, WPS-2-unknown-class, Dothideomycetes, and Glomeromycetes (Figure 3(c), Figure 3(d))).

In the homefield soils, heat stress reduced the relative abundances of the following bacteria taxa significantly: Acidobacteriia (from 1.3% relative abundance to 0%); α-proteobacteria (13.5% to 1.75%), γ-proteobacteria (14.51% - 0.05%), Bacteroidia (2.66 - 0.02), Blastocatellia (2.24% - 0.5%), Thermoleophilia (2.37% - 0.29%), WPS-2 (0.2% - 0.0%). The relative abundances of two fungi taxa also decreased: Dothideomycetes (25.68% - 7.25%) and Eurotiomycetes (8.43% - 3.67%). In contrast, the relative abundances of the following taxa increased significantly: the bacteria Actinobacteria (6.11% - 18.73%), Bacilli (42.08% - 56.58%), Chloroflexia (0.82% - 19.62%), and Clostridia (0.46% - 1.70%), as well the fungi taxa Sordariomycetes (44.41% - 69.92%) and Glomeromycetes (0.02% - 1.20%) (Figure 4(a), Figure 4(b)).

In the outfield soils, heat stress reduced the relative abundances of the following bacteria taxa significantly: Acidobacteriia (1.17% - 0.0%), Actinobacteria (10.50% - 0.11%), Chloroflexia (0.24% - 0%), Thermoleophilia (2.45% - 0%), WPS-2 (2.41% - 0%). The relative abundances of four fungi taxa also decreased: Chytridiomycetes (0.34% - 0%), Dothideomycetes (72.81% - 0.30%), Glomeromycetes (0.11% - 0%), and Sordariomycetes (8.09% - 0.27%) (Figure 4(a), Figure 4(b)). However, the relative abundances of the fungi taxa Eurotiomycetes increased significantly (13.82% - 91.84%), as did the abundances of the bacteria taxa α-Proteobacteria (10.35% - 26.74%), γ-Proteobacteria (11.10% - 14.32%), and Clostridia (0.6% - 0.78%), (Figure 4(a), Figure 4(b)). Comparisons between heat-stressed outfield soils samples and the controls found no significant change in the relative abundances of the most dominant Bacilli class (54.01% in the controls, versus 53.53% in heat-stressed samples). Bacilli in outfield soils seemed to be insensitive to heat stress (Figure 4(c), Figure 4(d)).

At the species level, we found that the following bacteria species dominated in the homefield control soils: Bacillus fumarioli (11.76%), Ammoniphilus resinae (5.81%), Microvirga sp (4.29%), and Flavisolibacter sp. (3.77%). Heat stress significantly reduced the relative abundance of Microvirga sp. and Flavisolibacter sp, but allowed a significant increase in the relative abundances of Sphaerobacter thermophilus, Aeromicrobium sp., and Cohnella sp (Figure 4(c)). The most dominant fungal species in the homefield control soils were Phaeoacremonium minimum (22.21%), Chaetomium sp. (21.49%), Pleosporales sp (8.36%), and Westerdykella cylindrica (5.20%). Heat stress reduced the relative abundances of those dominant species, but significantly increased the relative abundance of Sarocladium sp. (0.01% - 80.05%) (Figure 4(d)).

In the outfield control soils, the most dominant bacteria species were Bacillus fumarioli (24.46%), Paraburkholderia fungorum (8.33%), Tumebacillus sp. (6.90%), Arthrobacter sp. (7.28%), Ammoniphilus resinae (6.61%), and Sphingomonas echinoides (6.51%). Heat stress significantly reduced the relative abundances of Ammoniphilus resinae, Bacillus fumarioli, and Arthrobacter sp., while significantly increasing those of Tumebacillus sp., Sphingomonas echinoides, and Cohnelle sp. (Figure 4(c)). The most dominant fungal species in the outfield soils control were Boeremia exigua var. exigua (39.17%), Alternaria alternata (16.67%), Rhodotorula sp. (6.57%), and Penicillium chrysogenum (6.52%). Heat stress significantly reduced the dominant species, and allowed Aspergillus lentulus and Exophiala oligosperma to become dominant (Figure 4(d)).

Some but not all of the responses of the homefields’ microbial communities differed from those of outfields. In addition, some responses of heat-stressed soils differed from those of the controls. For example, heat-stressing of both soils produced a high biological activity that did not occur in the controls. However, the C mineralization in homefield soils was greater than in outfield soils. These results are consistent with those of Ågren and Wetterstedt [45] , who attribute them to a strong, temperature-induced C mineralization, and thus to an increase in basal respiration of the microbial community. Franco-Andreu et al. [46] explains the same results by citing mineralization of labile carbon an immediate source of energy for microorganisms.

For homefield soils as well as outfield soils, and for all heat-stress durations, C-CO₂ accumulations at the end of incubation were higher than those of the controls. As one example, the accumulations (taken as a class) for homefield soils that received the SD₂₈ heat stress exceeded those of the controls by 58.25 µg C-CO₂ g⁻¹ soil. However, accumulations for outfield soils that received SD₂₈ exceeded those of the controls by only 2.45 C-CO₂ g⁻¹ soil. Chotte et al. [1] obtained similar results in their research on manured and non-manured soils: the C-CO₂ accumulation of a heat-stressed manured soil exceeded that of the control by 68.8, µg C-CO₂ g⁻¹, versus 15.2 for the non-manured soil. That difference is explained by the high level of organic C in Chotte et al. [1] soils (20.1 and 18.3 mg·g⁻¹ soil, for manured and unmanured soils respectively). By comparison, the organic C contents in our soils were 10.5 and 3.9 mg·g⁻¹ soil in homefield and outfield samples, respectively.

Chotte et al. [1] work also agrees with our finding that C mineralization is significantly higher in soil amended with OM (homefield) than in the unamended (outfield). Specifically, Chotte et al. [1] found that the respiration of samples amended with OM is significantly higher than for unamended samples. Ågren and Wetterstedt [45] attribute the difference in C mineralization to temperature-induced solubilization of organic compounds.

The decrease in soil microbial respiration at the end of incubation is explained by a decrease in soil’s labile C fractions due to decomposition during heat stress [47] . Davet [48] agrees and states that biological activity decreases with resource depletion after 2 to 3 weeks at a level equal to or lower than that of the control soil. The slight recovery of microbial activity that occurred in our samples at the beginning of recovery confirms the results of Pailler [49] , who attributed it to a significant development of microbial metabolism. At the end of a disturbance (according to that author), the microbial community increases its basal respiration in order to resume metabolic activities.

C mineralization in homefield soils that underwent SD₃ and SD₁₄ was not statistically different from that of the controls. This result suggests that the microbial biomass resisted the SD₃ and SD₁₄ heat stresses. However, and in contrast to the raw-data comparison, the RS indices that we calculated from C-mineralization data did not show complete resistance to those two durations of heat stress. Furthermore, none of the RS values for homefield and outfield show complete resistance of the microbial biomass to any of the 3 durations. These findings are similar to those of Chotte et al. [1] , who found that OM inputs to homefield did not modify the microbial community’s resistance. Ben Sassi [13] also states that OM inputs do not improve microbial stability to an important degree. In contrast, Wada and Toyota [21] , show that the resistance of biological functions is higher in the presence of OM. The discrepancy between that result and our own could be explained by the organo-mineral amendment that was used on fields that were studied by Wada and Toyota.

In our own study, microbial communities in the OM-rich soil of homefield had high resistance to SD₃ and SD₁₄ heat stresses (RS = 0.70 and 0.80 respectively for SD₃ and SD₁₄). Therefore, those durations of heat stress do not disrupt all biological activities in the homefield samples. In contrast, RS indices for outfield soils were low for all three heat-stress durations. These results are similar to those of Griffiths et al. [50] , who found thermal-stress resistance in soils amended with OM. However, our results show that the SD₂₈ heat stress did disrupt microbial activity of homefield soils (RS = 0.01). Thus, the response of microorganisms to heat stress depends upon the duration of the stress [13] [14] .

The resilience indices (RL) of homefield and outfield soils show no recovery of the microbial biomass for any of the 3 durations of heat stress. This result is similar to those of Fierer and Schimel [51] , who observed no resilience after 6 weeks of recovery. The absence of complete resilience may be due to insufficient recovery time. For example, Kumar et al. [42] found that microorganisms needed 56 days to recover fully. Too, the extreme temperature of 60˚C can impede recovery by removing a large proportion of the active microorganisms from the soil [12] .

However, in the present study the RL indices do show a beginning of resilience in outfield soils after SD₃ (RL = 0.56) and SD₁₄ (RL = 0.76), and in homefield soils after SD₂₈ (RL = 0.63). Similarly Bérard et al. [31] found no complete resilience of microbial communities at the end of recovery. The partial recovery after SD₂₈ in homefield soils (versus the non-recovery in outfield soils) is explained by the organic amendment. Indeed, the soils richest in C are more resilient in response to intense heat stress [20] [21] [52] .

In our nutrient-poor outfield soils, heat stress not only decreased the microbial α-diversity significantly, but altered the abundances of most of the microbial taxa. However, only the fungal diversity decreased in the homefield soils. As will be explained below, these differences between the two soils may be attributable to several phenomena that interact in subtle ways. Those phenomena include the effects of heat upon the soils’ labile C contents, as well as the ranges of metabolic and functional flexibility within the two soils’ microbial communities.

In both soils, fungal and bacterial diversity was lower in the heat-stressed samples than in the controls. The decrease was greater in outfield soils. That result is consistent with Bécaert et al.’s [38] observation that heat stress can kill microbial communities, and that the impact tends to be greater in lower-OM soils. The high microbial diversity in the homefield soils is related to the legacy effect of OM, a substance that microorganisms depend upon to support their activity [8] .

In our study, OM’s legacy effect manifested itself in differences between the changes in taxonomic compositions of the two soils’ respective microbial communities. Heat stress reduced the relative abundances of homefield microbial communities that belong to α-proteobacteria, γ-proteobacteria, Bacteroidia, and Eurotiomycetes. In contrast, heat stress increased the abundances of those same microbes in outfield soils. One explanation for the decreased abundances in homefield soils may be that soil’s C resources diminished during heat stress, thereby exacerbating the vulnerability of the copiotrophic communities that normally proliferate in such resource-rich soils [53] [54] . The effect of heat stress upon C resources may also explain the outfield soils’ increased abundances of copiotrophs: heat stress probably promoted decomposition of recalcitrant SOM, thereby enriching soils with labile C. The increased abundance of copiotrophs in heat-stressed outfield soils is consistent with Davet [48] , finding that some species of Eurotiomycetes can grow at temperatures up to 60˚C [55] [56] .

An interesting contrast to that finding is offered by the relative abundances of Actinobacteria, Chloroflexia, and Sordariomycetes, which are highly stress-tolerant [57] . In the homefield soils, heat stress increased the abundances of these communities, perhaps by reducing competition from heat-sensitive microbes. In addition, the metabolic versatility of Actinobacteria, Chloroflexia, and Sordariomycetes enables them to develop in soils where recalcitrant carbon prevails [58] [59] [60] . For example, the Chloroflexia class is dominated by Sphaerobacter thermophilus species, whose optimal growth temperature is 55˚C - 60˚C [61] .

The above-mentioned metabolic versatility of microbes must be borne in mind when interpreting our RL indices, which were calculated from C-mineralization data (a measure of the microbial community’s functioning, rather than its taxonomic composition). As Preece et al. [62] notes, a partial resilience such as that which occurred in both of our soils may accrue from the microbial communities’ capacities for functional redundancy. Similarly, Riah-Anglet et al. [37] , found that heat stress does not affect bacterial abundances in soils that are rich in Actinobacteria and Bacteroidetes. For example, the variable and versatile physiology of Proteobacteria gives them a competitive advantage in various ecological niches [63] . As the continuing heat stress reduces competition from other species for soil resources [37] , new stress-resistant microbial communities with greater resilience and functional stability can develop [64] . In the event of a disturbance, these communities may either develop adaptation strategies (resistance) [65] , or remain inactive while waiting for conditions to become favourable (resilience) [66] .

In our study, the representatives from the most dominant bacterial class (Bacilli) seemed to be insensitive to heat stress, as evidenced by the fact that the relative abundance of that class did not decrease in the outfield soils. Nor did heat stress change the taxonomic composition of that class in outfield soils. Allison and Martiny [67] , define a microbial community’s composition as resistant if that composition is difficult to perturb. That resistance is enhanced by high degrees of metabolic flexibility and physiological tolerance to changing environmental conditions Allison and Martiny [67] . The heat-stress resistance of Bacilli in our study is consistent with previous studies stresses [53] [60] [68] that posit the production of heat-resistant endospores by Bacillus species as a means of resisting environmental stresses. As Kämpfer et al. [69] noted the Bacilli class is dominated by Cohnella sp, Lysinibacillus macrolides, Aeromicrobium sp., and Sphingomonas echinoides, which are able to grow at temperatures between 20 and 55˚C. In our present study, the Actinobacteria, Chloroflexia, and Sordariomycetes taxa (in homefield) and the Bacilli, α-proteobacteria, γ-proteobacteria, and Eurotiomycetes (in outfield) demonstrated their ability to resist heat stress via different strategies that members of those taxa employ to resist environmental stresses, and adapt to them.