This is the accepted version of the article: Preece, Catherine; Verbruggen, Erik; Liu, Lei; [et al.]. «Effects of past and current drought on the composition and diversity of soil microbial communi- ties». Soil biology and biochemistry, Vol. 131 (April 2019), p. 28-39. DOI 10.1016/j.soilbio.2018.12.022 This version is available at https://ddd.uab.cat/record/200861 under the terms of the license 1 Effects of past and current drought on the composition and diversity of soil 2 microbial communities 3 4 Catherine Preece1,2, Erik Verbruggen3, Lei Liu1,2, James T. Weedon4, Josep Peñuelas1,2. 5 1 6 CREAF, Cerdanyola del Vallès, 08193, Spain. 2 7 CSIC Global Ecology Unit, CREAF-CSIC-UAB, Bellaterra 08193, Spain; Department of Ecological 8 Science, Plants and Ecosystems. 3 9 Department of Biology, University of Antwerp, 2610 Wilrijk, Belgium. 4 10 Systems Ecology, Department of Ecological Sciences, Vrije Universiteit Amsterdam, 1081 11 HV, Amsterdam, The Netherlands. 12 13 Corresponding author: Catherine Preece – catherine.preece09@gmail.com 14 15 16 17 18 Keywords: drought, legacy effect, fungal:bacterial ratio, Mediterranean, Quercus ilex. This is the accepted version of the following article: Preece, C. et al. Effects of past and current drought on the composition and diversity of soil microbial communities in Soil biology and biochemistryt (Ed. Elsevier), vol. 131 (April 2019), p. 28-39. - Which has been published in final form at DOI 10.1016/j.soilbio.2018.12.022 © 2018. This manuscript version is made available under the CC-BY-NC-ND 4.0 license http://creativecommons.org/licenses/by-nc-nd/4.0/ 1 19 Abstract 20 Drought is well known to have strong effects on the composition and activity of soil 21 microbial communities, and may be determined by drought history and drought duration, but 22 the characterisation and prediction of these effects remains challenging. This is because soil 23 microbial communities that have previously been exposed to drought may change less in 24 response to subsequent drought events, due to the selection of drought-resistant taxa. We set 25 up a 10-level drought experiment to test the effect of water stress on the composition and 26 diversity of soil bacterial and fungal communities. We also investigated the effect of a 27 previous long-term drought on communities in soils with different historical precipitation 28 regimes. Saplings of the holm oak, Quercus ilex L., were included to assess the impact of 29 plant presence on the effects of the drought treatment. The composition and diversity of the 30 soil microbial communities were analysed using DNA amplicon sequencing of bacterial and 31 fungal markers and the measurement of phospholipid fatty acids. 32 The experimental drought affected the bacterial community much more than the 33 fungal community, decreasing alpha diversity and proportion of total biomass, whereas 34 fungal diversity tended to increase. The experimental drought altered the relative abundances 35 of specific taxa of both bacteria and fungi, and in many cases these effects were modified by 36 the presence of the plant and soil origin. Soils with a history of drought had higher overall 37 bacterial alpha diversity at the end of the experimental drought, presumably because of 38 adaptation of the bacterial community to drought conditions. However, some bacterial taxa 39 (e.g. Chloroflexi) and fungal functional groups (plant pathogens and saprotrophic yeasts) 40 decreased in abundance more in the pre-droughted soils. 41 Our results suggest that soil communities will not necessarily be able to maintain the 42 same functions during more extreme or more frequent future droughts, when functions are 2 43 influenced by community composition. Drought is likely to continue to affect community 44 composition, even in soils that are acclimated to it, tending to increase the proportion of fungi 45 and reduce the proportion and diversity of bacteria. 46 47 1. Introduction 48 Drought is a serious problem in many parts of the world, and the impacts on plants, in both 49 natural and agricultural settings, are increasingly well documented. The effects of water stress 50 on soils and their associated biota, however, remain less certain, even though water stress 51 may be the most frequent environmental stress experienced by soil microorganisms (Schimel 52 et al., 2007), with demonstrated impacts on soil properties and microbial communities. 53 Severe and prolonged water stress, in the most extreme scenarios, leads to desertification but 54 may have important consequences for soil health much sooner. An estimated one fifth of the 55 Earth’s soil is currently acutely degraded and showing declining productivity, often due to 56 drought-related phenomena (United Nations Convention to Combat Desertification, 2017). 57 Maintaining healthy soils is important because soils are fundamental for a wide range of 58 ecosystem services, including food security, nutrient cycling, timber production, and climatic 59 regulation. Moreover, the increasing research into belowground processes and plant-soil 60 interactions and the availability of methods for studying microbial communities have raised 61 interesting fundamental questions about the role of roots in mediating the effect of drought on 62 soil microbial communities. 63 Drought can directly affect microbes by desiccation or resource limitation, because 64 substrate diffusion is reduced at low levels of soil moisture (Schimel et al., 2007; Naylor and 65 Coleman-Derr, 2017). Drought experiments have reported decreases in microbial biomass 66 and activity (Hueso et al., 2012; Alster et al., 2013; Hartmann et al., 2017; Castaño et al., 3 67 2018), reductions in carbon and nitrogen mineralisation (Hueso et al., 2012), and 68 accumulation of solutes, such as amino acids (in bacteria) and polyols (in fungi), which help 69 prevent dehydration but are energetically expensive (Schimel et al., 2007). Drought may also 70 have indirect effects, through interactions with plants, because plants can have species- 71 specific effects on rhizosphere microbiota mediated by rhizodeposits (Bergsma-Vlami et al., 72 2005; Haichar et al., 2008; Bressan et al., 2009; Ladygina and Hedlund, 2010; Philippot et al., 73 2013; Lareen et al., 2016). Plants may have a protective effect on microbes that live in or near 74 the rhizosphere, at least when normal root function can be maintained. The tolerance of a 75 plant species to drought can therefore be important for the soil community in the immediate 76 vicinity since the presence of a tolerant plant may modulate the impacts on soil. 77 Soil microbial communities vary greatly at all geographical scales, depending on 78 factors such as the chemical properties of the soil, the climate, and the plant community, but 79 some general factors associated with drought have been identified. For example, evidence 80 suggests that fungi are more tolerant than bacteria to water stress (Bapiri et al., 2010; Barnard 81 et al., 2015; de Vries et al., 2018), and Gram-positive bacterial lineages are generally more 82 drought resistant than Gram-negative lineages, perhaps due to their thicker cell walls 83 (Schimel et al., 2007). Some evidence suggests that among fungi, yeasts may have a high 84 tolerance to drought, because they tend to be more common in more extreme environments 85 and tend to reproduce by budding, which is generally a more stress tolerant strategy of 86 reproduction (Treseder and Lennon, 2015). Water stress may affect the taxonomic diversity 87 of microbial communities, and both decreases (Bouskill et al., 2013) and increases (Acosta- 88 Martínez et al., 2014) have been reported. Microbial communities with high diversity, and 89 particularly high functional diversity, may be more tolerant to drought (and to other 90 perturbations), but this tolerance is likely to be strongly associated with a range of biotic and 4 91 abiotic features of the soil (Griffiths and Philippot, 2013). Many uncertainties about the 92 response of microbial communities to water stress, though, remain. 93 Both soil microbial communities and droughts are highly variable, in both natural and 94 experimental environments. The impact of water stress on plants and soils can depend on the 95 timing or duration of the event or treatment (Hoover and Rogers, 2016; Mengtian et al., 96 2018), the proportional change in water availability, and the historical precipitation regime 97 (Evans and Wallenstein, 2012; Bouskill et al., 2013). On this last point, soils that have 98 experienced drought events may demonstrate ‘legacy effects’ whereby the soil community 99 continues to show the impact of the drought for many years, and it may even modify the 100 response to a later drought (de Vries et al., 2012; Bouskill et al., 2013; Kaisermann et al., 101 2017; Meisner et al., 2018). The impact of drought may also not elicit a linear response 102 (Knapp et al., 2017), meaning that any negative effect may decrease or increase as the 103 drought progresses, indicating acclimation or a ‘tipping point’, respectively. All this 104 variability adds to the difficulty of drawing conclusions about the impacts of drought on soil 105 microbes, and experiments are needed to test a more complex range of drought scenarios, 106 including multiple levels of drought intensity in space and time and investigating if and when 107 biologically relevant thresholds are exceeded (Beier et al., 2012). 108 The main objective of this study was to investigate the effect of increasing drought 109 intensity on the bacterial and fungal communities of a Mediterranean soil. We focused on a 110 holm oak (Quercus ilex L.) forest system, which is a predominant habitat throughout the 111 Mediterranean Basin where greater drought frequency and severity is predicted (Field et al., 112 2014; Touma et al., 2015). We set up a greenhouse experiment with ten levels of drought and 113 then used DNA-based amplicon sequencing and lipid analyses to observe the effects on soil 114 bacterial and fungal communities in pots containing Q. ilex saplings. This experiment 115 allowed us to determine whether bacterial communities were more responsive than fungal 5 116 communities to drought and if we could identify changes in functional groups. We also 117 wanted to determine if the impacts of drought were mitigated or increased by the presence of 118 the plants and by the historical precipitation regime of the soil, and to determine the effect of 119 drought intensity. We hypothesised that: (1) drought would affect microbial community 120 composition, negatively affect diversity, and affect bacteria more than fungi, (2) the presence 121 of Q. ilex would decrease the impacts on the microbial communities, and (3) soils with a 122 history of drought would be more resistant to the drought treatment. 123 124 2. Materials and methods 125 2.1. Plant and soil material 126 A greenhouse experiment was established in May 2015 at the experimental fields of the 127 Autonomous University of Barcelona (Spain). The experiment comprised 180 pots of 3.5 l, 128 half of which (90 pots) were planted with three-year-old Q. ilex saplings (provided by 129 Forestal Catalana, Barcelona, Spain). The other 90 pots contained substrate only. The 130 substrate used in all pots consisted of 45% autoclaved peat (121oC for 60 mins), 45% sand, 131 and 10% natural soil inoculum. Soil was collected from a south-facing slope (25%) in a 132 natural holm oak forest in the Prades Mountains in northeastern Spain (41°13′N, 0°55′E; 930 133 m a.s.l.). This forest is the site of a long-term drought experiment that began in 1999 and 134 reduces precipitation throughfall by approximately 30% (Ogaya and Peñuelas, 2007). There 135 were three soil inocula, each with 60 replicates, which had different prior treatments: control, 136 pre-droughted, or pre-sterilised. We collected topsoil from the treatment plots of the long- 137 term drought experiment. Soil from the control plots was used as the inoculum for the 138 corresponding control soil in our experiment and was autoclaved (121oC for 60 mins) for use 139 in our pre-sterilised treatment, and soil from the drought plots was used as the inoculum in 6 140 our pre-droughted treatment. This pre-droughted soil allowed us to test for ‘legacy’ effects of 141 the long-term drought. The pre-sterilised treatment was incorporated in order to try to 142 separate the effects of the previous soil community from the drought and plant effects and 143 remove any potential idiosyncratic features of the historical soil community. The roots of the 144 Q. ilex saplings were carefully washed in tap water before transplantation to remove soil from 145 the previous potting mix, so that the soil communities were representative of the three new 146 soil treatments. Whilst complete removal of previous substrate and original rhizosphere 147 microbes was not possible without causing damage to the roots, the soil microbes in the bulk 148 soil would have been predominantly composed of those from the new inocula. All plants 149 were then allowed to adjust to the greenhouse environment for six weeks, receiving daily 150 watering (until the end of June 2015), after which they were top-watered every day with 151 amounts sufficient to maintain soil moisture at 20-25%. 152 153 2.2. Experimental design 154 The drought treatment consisted of ten levels of drought, applied by withholding water for 0, 155 2, 4, 7, 9, 11, 14, 16, 18, and 21 days. Each drought level therefore had 18 pots, divided into 156 six replicate blocks (Supplementary Material Fig. S1). Soil samples were collected at the end 157 of the drought period specific for each drought level. For pots with zero days of drought this 158 sampling occurred at the end of the six week acclimation period. A subset of 90 of these soil 159 samples were used for the DNA sequencing analysis, representing all drought levels and all 160 soil history and plant treatments. This included 13-15 replicates for each combination of soil 161 history × presence of Q. ilex. One sample was discarded for the bacteria and two for the fungi 162 (see Supplementary Material, Table S1 for full details of replication). Air temperature was 163 monitored throughout the experiment using an EL-USB-2 data logger (Lascar Electronics, 164 Wiltshire, UK) and had a mean of 26.7 °C. Soil temperature averaged 27.0 °C across the 7 165 three soil types (Decagon Em50 data logger with 5TM soil probes, Decagon Devices, 166 Pullman, USA) (Supplementary Material Fig. S2). Soil moisture in each pot was measured at 167 the start of the experiment and at the end of its drought period using an ML3 Theta Probe 168 connected to a HH2 Moisture Meter (Delta-T Devices, Cambridge, UK). Mean soil moisture 169 was 22.6% at the start of the experiment and decreased exponentially to 0.3% by the end of 170 the 21-day drought treatment, and this did not differ significantly between pots with plants 171 and those without (Supplementary Material Fig. S3). 172 173 2.3. DNA library preparation and sequencing 174 Total community DNA was extracted from approximately 0.25 g of soil using a PowerSoil 175 DNA Isolation Kit following the manufacturer’s protocol (MoBio, Carlsbad, USA). The 176 hypervariable V3-V4 regions of the bacterial 16S rRNA gene was amplified using the 341F – 177 806R primer pair (Klindworth et al., 2013) modified to include Illumina adapter sequences. 178 Each 25 μl reaction mixture contained 1.5 μl of undiluted DNA extract, 1 μl each of the 179 forward and reverse primers (10 μM), and 12.5 μl of Phusion High Fidelity PCR Master Mix 180 with HF Buffer (ThermoFisher Scientific, Waltham, USA). Initial DNA concentration ranged 181 between 3.9-10.2 ng/µl and DNA concentration was not standardized prior to PCR. PCR 182 conditions were: initial denaturation at 98 °C for 30 s, followed by 30 cycles of denaturation 183 at 98 °C for 30 s, annealing at 55 °C for 30 s, and extension at 72 °C for 30 s, with a final 184 extension at 72 °C for 4 min. After confirming successful amplification by agarose gel 185 electrophoresis, PCR products were purified and normalized using Sequalprep plates 186 (Thermofisher, USA), and subject to a second indexing PCR such that each sample received a 187 unique combination of 6-nucleotide barcoded forward and reverse primers. The reaction 188 mixture was as above, and the PCR program was an initial step at 95°C for 30 s, 8 cycles of: 189 95°C for 30 s, 55°C for 30 s, 72°C for 30 s; and a final step of 72°C for 5 min. PCR products 8 190 were again purified and normalized with Sequalprep plates and pooled for sequencing. The 191 fungal ITS1 region was amplified in a one step approach using the primers ITS1f and ITS2 192 augmented with multiplexing barcodes (Smith and Peay, 2014). Each reaction mixture 193 contained 1 µl of the DNA extract, 1 μl of forward and reverse primers (10 μM), 200 μM 194 dNTP's, 1X GC buffer and 0.4 U of Phusion DNA polymerase. PCR conditions were: initial 195 denaturation at 98 °C for 30 s, followed by 40 cycles of 98 °C for 30 s, 55 °C for 30 s, and 72 196 °C for 30 s, and a final extension at 72 °C for 10 min. Samples that failed to produce a PCR 197 product were discarded, and PCR was repeated. Some fungal samples still failed to produce 198 usable PCR products and were excluded from further analyses, but this only represented three 199 samples out of 88 in total. Fungal PCR products were also purified and normalized with 200 Sequalprep plates, and additionally extracted from a 1.5 agarose gel for size selection 201 (approximately 200-500 bp which covers the entire range of length variation in the fungal 202 ITS1 region) and to remove primer dimers. Then they were additionally purified using a 203 QIAquick Gel Extraction Kit (Qiagen, Venlo, The Netherlands). Both fungal and bacterial 204 libraries were quantified with real-time PCR using KAPA Library Quantification Kits (Kapa 205 Biosystems, Wilmington, USA) to determine dilution factors for the sequencing protocol. 206 The libraries were sequenced on the Illumina MiSeq platform (Illumina Inc., San 207 Diego, USA), with 2 x 300 cycles (V3 chemistry) for forward and reverse reads for bacteria, 208 and 300 cycles (V2 chemistry) in the forward direction only for fungi. The reproducibility of 209 sample preparation and sequencing was tested by sequencing a small number of technical 210 replicates (DNA isolated from the same samples but subjected to independent PCR reactions 211 with distinct primer barcodes). 212 213 214 9 215 2.4. Quality filtering and bioinformatic analysis 216 The initial bioinformatic analysis of bacterial sequences used USEARCH software (Edgar, 217 2013). Merging of paired-end reads was attempted, but low quality basecalls in the tail of 218 both read directions precluded satisfactory merging. We therefore proceeded with separate 219 parallel analyses of the forward and reverse reads (after truncating to 180bp and 150 bp 220 respectively). Both reads gave qualitatively very similar results, so we focus hereafter on the 221 longer forward read data. After primer removal sequences were filtered by quality leaving a 222 total of 2.4 M high-quality sequences. Replicate and singleton sequences were removed, and 223 a set of representative sequences of operational taxonomic units (OTUs) (97% similarity) was 224 constructed using the UPARSE-OTU algorithm (Edgar, 2013). Chimeras were removed 225 (leaving 11890 non-chimeric OTUs), and all original reads were mapped to the non-chimeric 226 OTUs using the USEARCH algorithm with global alignments at an identity threshold of 0.97, 227 yielding an OTU table. All subsequent steps used QIIME (Caporaso et al., 2010b). OTUs 228 were aligned using the PyNAST algorithm (Caporaso et al., 2010a) and the Green Genes 229 database (release 13_8, DeSantis et al., 2006) as a template alignment. A subsampled OTU 230 table was created by randomly sampling the original OTU table to avoid artefacts associated 231 with library size. Samples that contained fewer sequences than the requested depth (2480) 232 were omitted from the output OTU tables. Each OTU was taxonomically identified based on 233 the 97% Green Genes database (release 13_8) using the RDP classifier (Wang et al., 2007). 234 Fungal sequences were analysed using USEARCH following the UPARSE pipeline 235 (Edgar, 2013). The sequences were trimmed to 250 bp and filtered for quality (maximum 236 expected error of 0.5), leaving a total of 9.79M sequences. While we only sequenced the 237 forward reads, in some instances (sequences shorter than 250 bp) the reverse primer is found 238 at the end of the forward reads and these were removed. Then, Ns were added up to 250 bp 239 for efficient clustering of the OTUs. Singleton sequences were removed, and all others were 10 240 clustered to 97% similarity. Chimeras were filtered de novo and through the UNITE database 241 of ITS1 sequences implemented in UCHIME, leaving a total of 3,323 non-chimeric OTUs, 242 after which original sequences were mapped against these OTUs at a similarity threshold of 243 97% and assembled in an OTU table. Representative sequences for each OTU were aligned 244 to all fungal representative species in the UNITE database (Kõljalg et al., 2005) (release date 245 20.11.2016) using the BLAST algorithm with default settings. The resulting hits were 246 assigned to taxa, selecting the hit with the lowest E-value, provided it had a minimum E- 247 value of 1x10-36 and a minimum alignment length of 75 bp. OTUs were subsequently 248 assigned to functional groups if a genus was provided for the highest hit, and if it matched 249 one of the genera with known lifestyles provided by Tedersoo et al. (2014). When the genus 250 level was unknown, lifestyle was assigned at the family level if >80% of the genera within 251 that family (represented by more than three genera) had the same lifestyle. As for bacteria, 252 the OTU table was downsampled, to 12,332 reads per sample. 253 254 2.5. Analysis of phospholipid fatty acids 255 We used the amounts of phospholipid fatty acids (PLFAs) to quantify microbial biomass and 256 the ratio of fungal to bacterial biomass. PLFA extraction and identification followed 257 Frostegård et al. (1993), using 1 g of freeze-dried soil from each replicate. The abundance of 258 individual fatty acids was determined as nmol per g of dry soil, and standard nomenclature 259 was used (Tunlid et al., 1989). Concentrations of each PLFA were calculated based on the 260 19:0 internal standard concentrations. Selection of bacterial PLFAs follows the selection of 261 fatty acids by Frostegård and Bååth (1996), of which i14:0, i15:0, a15:0, i16:0, 16:17c, 262 a17:0, i17:0, cy17:0, 18:17, and cy19:0 were present in our samples. We calculated the sum 263 of i14:0, i15:0, a15:0, i16:0, a17:0, and i17:0 as an indicator of Gram-positive bacteria. 264 Gram-negative bacteria were identified by the PLFAs 16:17c, cy17:0, 18:17, and cy19:0 11 265 (Zelles, 1999). The fungi were identified by 18:26 (Frostegård et al., 1993; Frostegård et al., 266 2011). The ratio of 18:26 to total bacterial PLFAs was used to estimate the ratio of fungal to 267 bacterial biomass in soils (Bardgett et al., 1996; Frostegård and Bååth, 1996). 268 269 2.6. Statistical analyses 270 All statistical analyses were carried out using R v3.4.1 (R Core Team, 2016). The data for the 271 fungal and bacterial communities were log-transformed relative abundance of each OTU 272 within a sample. Non-metric multidimensional scaling (NMDS) plots (using Bray-Curtis 273 distance with the metaMDS function in vegan) were constructed to visualize the separation 274 between treatments. A permutational multivariate analysis of variance (PERMANOVA) was 275 used to assess the effects of drought, soil history, and presence of a plant on the microbial 276 communities, using the adonis2 function in the vegan package. Block was included as a 277 random factor, and 3000 permutations were used. This used Bray-Curtis dissimilarity indices 278 generated by the vegdist function (also from vegan). A pair-wise PERMANOVA analysis 279 was used to test for significant differences between pairs of factor levels (Martinez Arbizu, 280 2017). We also carried out an NMDS analysis and PERMANOVA with the pre-sterilised soil 281 removed, to check that this sterilised soil was not driving all of the significant soil history 282 effect, and of the non-droughted pots only to assess the differences in the soil communities 283 under ambient water conditions. 284 Shannon diversity (H) and Simpson index were calculated as a measure of community 285 alpha diversity (later referred to simply as community diversity) using the diversity function 286 (in vegan) and analysed using linear mixed effects models (lme function in the nlme package) 287 with drought intensity, soil history and plant-presence as predictor variables, and block as a 288 random factor. The change in community diversity following drought (H) was calculated 289 for each soil and plant treatment as (Hdmax – Hcon)/Hcon *100, where Hcon is the mean Shannon 12 290 diversity under normal water conditions (zero days of drought), and Hdmax is the mean 291 Shannon diversity under the most extreme drought level (21 days of drought). Unfortunately, 292 due to a lack of paired samples for Hcon and Hdmax, only means per treatment combination 293 could be calculated, and not standard errors. Also, a two sample t-test was done to compare 294 the mean Shannon diversity of the bacteria and fungi communities under control conditions 295 compared with the most extreme drought. 296 The relative abundance of bacterial phyla and classes with more than 2% mean 297 relative abundance, and fungal functional groups with more than 1.5% mean relative 298 abundance was analysed using linear regression, with drought intensity, soil history and 299 plant-presence as predictor variables. PLFA data indicating total microbial abundance was 300 analysed using a linear mixed-effects model (lme function in the nlme package) with block as 301 a random factor. The fungal:bacterial ratio and the ratio of Gram-positive to Gram-negative 302 bacteria were analysed with a generalised linear model with a quasi-binomial distribution. 303 In all analyses except relative abundance of taxa, drought was analysed both as a 304 continuous and a categorical (non-ordinal) variable with levels grouped as control (0 days of 305 drought), low-level drought (2-7 days of drought), mid-level drought (9-14 days of drought), 306 and high-level drought (16-21 days of drought). These groups were selected in order to 307 equally divide the drought treatments, each spanning a 6-day period. Results are usually 308 shown with drought as a continuous variable, unless otherwise stated, in which case it is due 309 to a non-linear relationship between the dependent variable and drought, specifically for 310 Shannon diversity for fungi. In all cases, the P-values shown are the result of an ANOVA 311 (Anova function in the car package) with type III sums of squares. 312 313 314 13 315 3. Results 316 3.1. Initial differences in the soil communities 317 The NMDS analysis suggested that soil history and the presence of the plant affected the 318 composition of the bacterial community before the start of the drought treatment (after six 319 weeks of acclimation) (Fig. 1a, Fig. 1b, Table 1; PERMANOVA, soil history effect, P < 320 0.001, pseudo R2 = 0.25, plant effect, P < 0.001, pseudo R2 = 0.16). Note that the soil effect 321 was driven by the strong difference between the pre-sterilised soil and the control and pre- 322 droughted soils (Pairwise PERMANOVA, P < 0.05), with no significant difference between 323 the latter two soils. The presence of Q. ilex did not affect the fungal community (Fig. 1c), but 324 soil history did, with the pre-sterilised soil clearly separated from the control and pre- 325 droughted soils (Fig. 1d; PERMANOVA, soil history effect, P < 0.001, pseudo R2 = 0.51). 326 The diversity of the bacterial community measured by the Shannon index was 327 affected by an interaction between the soil history effect and the presence of the plant (Table 328 1; significant interaction, χ² = 29.5, P < 0.001) such that the pre-sterilised soil had lower 329 diversity than the other soils, but only in the absence of the plant (Supplementary Material, 330 Fig. S4a). The Simpson index showed a very similar pattern, again with a significant 331 interaction between plant and soil history (Table 1; significant interaction, χ² = 21.9, P < 332 0.001), again with the lowest value for pre-sterilised soil without the plant (Supplementary 333 Material, Fig. S5a). 334 Soil history, but not the plant, affected the diversity of the fungal community. 335 Diversity of the fungal community was higher in the pre-droughted than the pre-sterilised soil 336 (Table 1; χ² = 11.9, P < 0.01), with the control soil intermediate, and the diversity was 14.5% 337 higher in the pre-droughted than the control soil, but not significantly higher (Supplementary 338 Material, Fig. S4b). The Simpson index showed a similar pattern (Supplementary Material, 14 339 Fig. S5b) although with an interaction between soil history and plant (Table 1; χ² = 9.3, P < 340 0.01), so that the pre-droughted soil without Q. ilex had lower diversity than the control soil 341 with Q. ilex present. 342 343 3.2. Microbial abundance 344 The fungal:bacterial ratio increased with experimental drought duration (χ² = 10.1, P < 0.01) 345 and was also higher in pots containing only soil than those containing Q. ilex (χ² = 34.2, P < 346 0.001) (Fig. 2a). Soil history had no effect on the fungal:bacterial ratio. Drought had no effect 347 on total microbial biomass, but soil history significantly interacted with the presence of Q. 348 ilex in the pots containing only soil (χ² = 24.3, P < 0.001). In the absence of Q. ilex, microbial 349 biomass was higher in the control and pre-droughted soils than the pre-sterilised soil. 350 However, microbial biomass was lower in the pre-droughted than the control soils when Q. 351 ilex was present, with intermediate biomass in the pre-sterilised soil (Fig. 2b). The ratio of 352 Gram-positive:Gram-negative bacteria decreased as drought duration increased (χ² = 5.5, P < 353 0.05) and was affected by soil history (χ² = 21.3, P < 0.001), being lower in the pre-sterilised 354 soil and in soils without Q. ilex (χ² = 11.5, P < 0.001) (Supplementary Material Fig. S6). 355 356 3.3. Composition of the microbial community during the drought experiment 357 Drought had a strong impact on the bacterial community (P < 0.001) but not the fungal 358 community (Table 2, Fig. 3). The result was almost identical when drought was grouped into 359 the four categories (control and low-, mid-, and high-level drought, Table 2), except for an 360 interaction between drought and the presence of Q. ilex for the bacterial community (Table 2; 361 P < 0.05). Although, note that the analysis without the pre-sterilised soil revealed an effect of 362 drought (P < 0.05) on the fungal community when drought was considered as a categorical 363 variable (Supplementary Material, Table S2). Differences between each pair of categorical 15 364 drought levels are shown in Supplementary Material (Table S3). Soil history was a significant 365 driver of both the bacterial and fungal communities (P < 0.001), and there were significant 366 differences (Pairwise PERMANOVA, P < 0.01) between all pairs of the three groups of soil 367 history (Table S4). This significant effect of soil history was maintained even when the pre- 368 sterilised soil treatment was removed from the analysis (Supplementary Material, Table S2), 369 however note that the R2 value decreases from 0.17 to 0.05, suggesting that the pre-sterilised 370 soil drives much of this effect. The presence of Q. ilex was important for bacteria (P < 0.001) 371 and fungi (P < 0.001). Soil history also influenced the effect of plant presence (there was a 372 significant interaction) on both bacteria and fungi community composition (Table 2, Fig. 3), 373 although for fungi this interaction was not seen when the pre-sterilised soil was removed 374 from the analysis (Supplementary Material, Table S2). 375 376 3.4. Diversity of the microbial community during the drought experiment 377 Drought duration had a negative effect on the diversity of the bacterial community (Fig. 4a, 378 Fig. S7a; Shannon diversity, χ² = 9.0, P < 0.01, Simpson index, χ² = 6.2, P < 0.05) but with 379 no interaction with either soil history or the presence of Q. ilex (Table 3). The three soils also 380 had different diversities, in the order pre-droughted > control > pre-sterilised (Shannon 381 diversity, χ² =12.7, P < 0.01), and the effect of soil history was modified by the presence of a 382 plant, with a strong plant effect increasing diversity in the pre-sterilised soil (Table 3, Fig.4b, 383 Fig. S7b; Shannon diversity, χ² =37.6, P < 0.001, Simpson index, χ² =19.4, P <0.001). 384 385 Drought had a positive effect on fungal Shannon diversity as a continuous variable (χ² = 3.9, 386 P < 0.05), but this effect was even greater when drought was separated into discrete groups 387 (control and low-, mid-, and high-level droughts) (χ² = 21.6, P < 0.001), with the low- and 16 388 high-level groups having higher Shannon diversities than the control group (Fig. 5a). This 389 same pattern was shown for Simpson index (Table 3, Fig. S8a; χ² = 19.0, P < 0.001). Drought 390 duration did not interact with either the presence of Q. ilex or soil history for fungi (Table 3). 391 Fungal diversity was generally lower in the pre-sterilised soil than the control and pre- 392 droughted soils (Shannon diversity, χ² =15.4, P < 0.01), and soil history interacted with the 393 presence of Q. ilex, with a larger positive effect of Q. ilex in the pre-sterilised soil (Fig. 5b, 394 Fig. S8b; Shannon diversity, χ² =9.0, P < 0.05, and Simpson index, χ² =13.6, P < 0.01). 395 For bacterial communities, all combinations of soil history and plant-presence showed 396 a decline in Shannon diversity between the control (0 days drought) treatment and the most 397 extreme (21 days) drought treatment (a negative H). Overall, for bacteria, a t-test on H in 398 the control and the extreme drought treatments showed that the difference between Shannon 399 diversity in the control and most extreme drought treatment was not significantly different 400 (Table 4; t = 1.72, P = 0.10). In contrast, fungal Shannon diversity tended to increase between 401 the most extreme drought treatment and the control drought level (a positive H), by up to 402 35% in one case, although in the case of soil with a history of previous drought and with 403 presence of Q. ilex Shannon diversity decreased slightly. Overall, for fungi, a t-test did not 404 reveal a significant difference between the two groups (control – 0 days drought – and most 405 extreme drought – 21 days drought) (t = -1.88, P-value = 0.07). 406 407 3.5. Microbial taxonomic composition 408 Overall, the most abundant bacterial phyla were Proteobacteria (36.0% of amplicon reads), 409 Actinobacteria (18.5%), Bacteroidetes (13.4%), and Verrucomicrobia (6.4%). Other phyla 410 that comprised a substantial (>2%) amount of the bacterial community were Planctomycetes 411 (5.0%), Chloroflexi (4.8%), Acidobacteria (4.5%), and Firmicutes (2.9%). Drought affected 17 412 most of these phyla (Table S5), and there were often interactions with either the plant- 413 presence or soil history. Actinobacteria and Planctomycetes abundance increased with 414 drought whereas Proteobacteria abundance was negatively correlated with drought duration. 415 For Bacteroidetes there was a negative correlation with drought in control soil, and for 416 Chloroflexi abundance increased with drought in control soil but decreased with drought in 417 pre-droughted soil. 418 Twelve bacterial classes were present at >2%: Alphaproteobacteria (17.5%), 419 Actinobacteria (11.4%), Betaproteobacteria (6.9%), Gammaproteobacteria (6.4%), 420 Saprospirae (5.6%), Thermoleophilia (5.5%), Cytophagia (5.3%), Deltaproteobacteria 421 (5.1%), Planctomycetia (3.3%), Opitutae (3.0%), Bacilli (2.8%), and Anaerolineae (2.6%). 422 Again, the drought treatment affected most classes, but the effect was highly variable and 423 depended on the presence of Q. ilex and soil history (Table S6). A large proportion of the 424 fungal OTUs (45.8%) were unidentified, but filamentous saprotrophic fungi were the most 425 abundant functional group (48.3%), followed by plant pathogens (2.5%) and saprotrophic 426 yeasts (1.9%) (Table S7). Other functional groups at lower proportions (< 1.5%) of the total 427 abundance included mycoparasites, ectomycorrhizal fungi, and saprotrophic white-rot fungi. 428 429 4. Discussion 430 The impact of various drought scenarios on soil and rhizosphere microbes remains uncertain 431 and is likely to depend on soil and plant properties. Few previous studies have measured the 432 response of soil microbial communities to more than two levels of drought, a knowledge gap, 433 which our study aimed to address. The experimental design allowed us to determine the 434 effect of drought on microbial communities in detail, both as a continuous variable, to 435 observe general trends, and grouping the drought levels, to increase replication in separate 18 436 groups and allow comparisons between drought intensities. Our results have demonstrated 437 that bacteria and fungi can have complex responses to water stress that vary with the intensity 438 of the drought as well as soil history and presence or absence of plants. 439 440 4.1. Differences in the soil communities without drought 441 Our analysis of pots in the zero days of drought (control) treatment, which were sampled after 442 six weeks of acclimation in the greenhouse, revealed that Quercus ilex plants impacted 443 bacterial community composition and had a positive effect on bacterial diversity but did not 444 affect the composition or diversity of the fungal community. This result may indicate either a 445 larger influence of the rhizosphere (e.g. root exudation) on bacteria than fungi, or that fungi 446 community composition responds more slowly due to their longer generation times (Rousk 447 and Bååth, 2011). Consistent with this supposed relative inertia of fungal communities, 448 previous soil history was found to have had a strong influence on the fungal community, 449 which may have masked any effect of the plant. Indeed, soil history clearly separated both the 450 bacterial and fungal communities in the sterilised soil from those in the control and pre- 451 droughted soils. This showed that the sterilisation treatment successfully removed most of the 452 bacteria and fungi present in the soil, and that the new microbial communities that colonised 453 the soil were different than the original ones. Soil history also affected the diversity of 454 bacteria and fungi in pots with zero days of drought. Bacterial community diversity was 455 higher in pre-droughted and control soils compared with sterilised soil, but only in the 456 absence of Q. ilex, indicating that the presence of a plant can quickly restore bacterial 457 diversity. This is supported by previous studies that have shown the importance of plants for 458 increasing bacterial diversity and richness, such as during habitat restoration after 19 459 contaminated soil (Yin et al., 2000; Harris, 2003). Fungal diversity was higher for the pre- 460 droughted than the pre-sterilised soil (but did not differ from the control). 461 462 4.2. Impact of drought on microbial-community composition 463 The drought treatment generally had no impact on total microbial biomass, which may seem 464 surprising, but previous studies have also reported mixed results. For example, an 465 experimental drought treatment actually increased microbial biomass in a mountain meadow 466 (Fuchslueger et al., 2014), hypothesised to be due to the continuation of carbon inputs from 467 plants during the drought, especially to fungi. Similarly, a study of two grasses and one 468 leguminous species recorded higher microbial biomass under drought when plants were 469 grown in mixtures and variable trends under monocultures (Sanaullah et al., 2011), and a 10- 470 month throughfall-exclusion experiment in a tropical forest found no effect on microbial 471 biomass (Bouskill et al., 2013). Others studies, however, have reported a decrease in 472 microbial biomass linked to lower soil water content, such as in a six-year drought 473 experiment in a semiarid forest (Bastida et al., 2017), in a short-term experiment without 474 plants (Chowdhury et al., 2011), and in a hardwood forest correlating natural variation in soil 475 moisture with soil microbial biomass (Baldrian et al., 2010). Importantly, total microbial 476 biomass in our study was lower in the pre-droughted than the control soils when the plant was 477 present, providing evidence that the long-term drought conditions damaged the soil microbial 478 community, rather than led to acclimation as is sometimes hypothesised. Soils with an 479 associated plant community represent a more realistic scenario than a bare, non-vegetated 480 soil, and this result may indicate that only a small proportion of the soil microbiota are able to 481 adapt to the drought conditions (Kaisermann et al., 2017). 20 482 Drought duration had a significant effect on the fungal:bacterial ratio, which increased 483 with increasing drought. This result is broadly consistent with the majority of previous 484 studies on this topic (Bapiri et al., 2010; Barnard et al., 2015). It is likely attributable to the 485 chitinous cell walls of fungi, which should increase their resistance to environmental 486 fluctuations, such as water stress (Holland and Coleman, 1987), and fungal hyphal growth 487 (which most bacteria do not have) allowing them to cross small areas of dry soil (Yuste et al., 488 2011). Trends in fungal versus bacterial dominance, however, are variable and may depend 489 on the trait measured (e.g. biomass or growth) and the method used (Strickland and Rousk, 490 2010). Drought also decreased the ratio of Gram-positive:Gram-negative bacteria, which was 491 surprising because Gram-positive bacteria are typically more drought resistant (Schimel et 492 al., 2007). This trait, however, may be linked to the increase in root exudation by Q. ilex 493 during increasing drought (Preece et al., 2018), because Gram-negative bacteria preferentially 494 consume this type of labile carbon source, whereas Gram-positive bacteria tend to consume 495 more recalcitrant C sources (Balasooriya et al., 2014; Naylor and Coleman-Derr, 2017). 496 In addition to reducing the proportion of bacterial biomass compared to fungi, drought 497 also clearly affected the composition of the bacterial community. Drought increased the 498 relative abundance of Actinobacteria, supporting previous studies that also reported this same 499 pattern in a range of soils and in the rhizosphere and endosphere of various plant species 500 (Bouskill et al., 2013; Nessner Kavamura et al., 2013; Naylor and Coleman-Derr, 2017). This 501 increase in abundance may be due to the ability of Actinobacteria to form spores, which 502 would allow them enter a dormant state during periods of environmental stress, such as 503 drought (Naylor and Coleman-Derr, 2017; Taketani et al., 2017). 504 The comparison of community compositions did not reveal an overall impact of 505 drought on fungal-community composition, but drought affected specific functional groups, 506 although these effects depended on soil history. For example, the relative abundance of plant 21 507 pathogens increased during drought in the control soils but decreased in the pre-droughted 508 and pre-sterilised soils. Phytophthora diseases such as P. cinnamomi are generally favoured 509 on soils where drainage is impeded (Desprez-Loustau et al., 2006) and have been shown to 510 increase in soils during adverse climatic conditions (such as drought or waterlogging) due to 511 host plants becoming less stress resistant, allowing a build up of pathogens in the soil 512 (Brasier, 1996; de Sampaio e Paiva Camilo-Alves et al., 2013). However, contrary to our 513 findings, most soil pathogens are thought to be favoured by wetter soils (Cook and 514 Papendick, 1972). 515 In addition, the relative abundance of yeasts was positively correlated with drought 516 duration in the pre-sterilised soil but decreased in the pre-droughted soil and did not change 517 in the control soil. This result suggests a complex network of interactions between 518 environmental factors (in this case water stress) and the pre-existing soil community, where a 519 difference in the history of the soil can generate variation in the subsequent response of 520 fungal community composition to perturbation. Previous research has predicted a potential 521 increase in yeasts under future drought scenarios, as they tend to be found in more stressful 522 environments (Treseder and Lennon, 2015). 523 524 4.3. Impact of drought on microbial-community diversity 525 The experimental drought generally had a negative effect on bacterial-community alpha 526 diversity (Shannon H and Simpson index), which has also been reported in a previous 527 drought study (Bouskill et al., 2013), although diversity is generally not affected (Bachar et 528 al., 2010; Acosta-Martínez et al., 2014; Naylor and Coleman-Derr, 2017; Tóth et al., 2017). 529 Drought had positive effects on fungal diversity, specifically in the low- and high-level 530 drought treatments compared to the control. Previous studies have reported higher fungal 22 531 diversity under drought (Acosta-Martínez et al., 2014; Schmidt et al., 2018) and may indicate 532 a higher tolerance of these organisms to drought, which would allow them to thrive if bacteria 533 are negatively affected. 534 In addition to the impact of drought on diversity, we noticed a clear pattern in the 535 effect of plants and soil. Not surprisingly, both bacterial and fungal diversity were lower in 536 the pre-sterilised soil than the other two soil histories, as communities in pre-sterilised soil 537 were much more recently created, with only a short time for colonisation of microorganisms. 538 Bacterial diversity was higher in the pre-droughted soil, indicating a positive effect of the 539 past precipitation regime. This agrees with a previous study that also found a positive legacy 540 effect of previous drought on bacterial diversity in a tropical forest soil (Bouskill et al., 2013). 541 The presence of the plant was very beneficial in augmenting both bacterial and fungal 542 diversity in the soil with the lowest diversity (pre-sterilised soil). This was almost certainly 543 due to the presence of microbes on the roots of the saplings at planting, which would have 544 had much more of an impact in this soil history type. Plant presence has been shown to have 545 beneficial effects on bacterial diversity, such as in a previous study in a semiarid shrub 546 system which found higher Shannon index, evenness and richness under the two shrub 547 species than under bare soil (Hortal et al., 2015). 548 When interpreting the effects of drought duration on the measured responses it is 549 important to note that samples for different drought duration treatments were taken at 550 different time points. It is therefore possible that underlying temporal dynamics, perhaps 551 related to disturbance during soil sampling, and therefore unrelated to the treatment, could be 552 causing the observed patterns. However, given the extended pre-incubation and stabilization 553 period of six weeks prior to the initiation of drought, and the fact that disturbance effects on 554 microbial community in incubations attenuate over relatively short time periods (e.g. Weedon 555 et al., 2013) we consider it unlikely that such temporal dynamics would be as large as drought 23 556 effects. It is also important to note that the interpretation of alpha diversity measures from 557 amplicon data can be problematic due to potentially spurious OTUs and the possibility of 558 some taxa falling under the detection limit due to incomplete sampling. These results should 559 therefore be considered as preliminary and be used as the basis for more detailed future 560 studies. 561 562 4.4. Are bacteria more affected than fungi by drought? 563 Drought tended to affect the community composition of bacteria more than fungi, and the 564 proportion of bacterial biomass compared with fungal biomass decreased under water stress, 565 even though total microbial biomass was unaffected by drought. Taken together, this could 566 suggest important changes in the future functioning of the bacterial community in soils 567 exposed to water stress, for example relating to carbon and nutrient cycling (Schimel et al., 568 2007; Frank et al., 2015). During drought (in water-limited areas) decomposition rates slow, 569 leading to a build up in soil organic matter (SOM) and lower N mineralisation (Borken and 570 Matzner, 2009; Larsen et al., 2011; van der Molen et al., 2011; Sanaullah et al., 2012; 571 Nguyen et al., 2018). The lack of strong effects of drought on fungal-community composition 572 may be due to different responses being found depending on the strength of the drought, thus 573 a lack of a clear unidirectional pattern: fungal diversity increased under low- and high-level 574 drought but not mid-level drought. Soil bacteria are more abundant than fungi, so fungi may 575 tend to take advantage of gaps where and when they can, resulting in a less standardised and 576 more idiosyncratic response to disturbance. For example, we could speculate that under low 577 drought there is some release of competition with bacteria, but during mid-level drought this 578 is offset by inhibitive effects, such as low substrate diffusion rates and energetically 579 expensive solute accumulation (Schimel et al., 2007). During extreme drought, some fungi 24 580 will benefit from being able to consume necromass, and this may especially favour fungi if 581 there is an increase in the C:N ratio of this dead mass (Moore et al., 2004) which can happen 582 under drought (Crowther et al., 2015). In addition, the bacterial community is able to respond 583 much quicker than fungi to the experimental conditions, thus it could be that bacteria are also 584 more rapidly affected by the imposed drought, and such responses are slower to be seen in 585 fungi. This would fit with previous studies that have found that soil bacterial activity was 586 more responsive to soil water content than fungal activity (Bell et al., 2008) and that changes 587 to bacterial communities under drought were longer lasting than for fungal communities (de 588 Vries et al., 2018). However, it is difficult to generalise across other systems, and there is 589 some evidence indicating that fungi are more sensitive than bacteria to smaller changes in soil 590 moisture (Kaisermann et al., 2015) with some cases where fungal abundance was more 591 greatly reduced by drought (Cregger et al., 2012). 592 593 4.5. Does the presence of Q. ilex lessen the impacts of drought on microbial 594 communities? 595 Previous studies have reported strong links between above and belowground communities, 596 and high plant diversity tends to increase soil microbial biomass and activity due to inputs of 597 organic matter and the regulation of soil moisture (Zak et al., 2003; Lange et al., 2015; 598 Thakur et al., 2015). The presence of plants can also shield microbial communities from the 599 impacts of drought (de Vries et al., 2012) and deep roots may act as moisture hotspots during 600 dry seasons (Castaño et al., 2018). We hypothesised that the presence of Q. ilex would be 601 beneficial for microbes (e.g. total microbial biomass), because plants are a source of carbon 602 inputs, such as root exudates and litter that provide a readily available energy source for 603 many microorganisms (Dennis et al., 2010). Indeed, a recent study of root exudation by our 25 604 study species, Q. ilex, under the same experimental conditions found that the exudation of 605 carbon increased during drought (Preece et al., 2018). Additionally, a review summarising 606 published drought impacts on root exudation in a range of species and with various 607 experimental methods found that carbon inputs tend to increase under drought, although this 608 effect may decrease or reverse under very severe drought (Preece and Peñuelas, 2016). The 609 impact of water stress on the overall composition and diversity of both bacteria and fungi in 610 our current study, however, was not affected by the presence of Q. ilex. Whilst we must be 611 careful about the conclusions drawn from amplicon data about alpha diversity, these results 612 suggest that any protective effect of the presence of the plant on microbial diversity would be 613 very minor, especially in comparison to other factors such as water stress or soil history. 614 Water is so limiting to microbial growth (Manzoni et al., 2011) that any positive 615 effect of the plant under drought conditions of more than a few days was likely minimal. 616 Although without measurements of water potential this is hard to confirm. An increase in 617 carbon inputs to the soil by plants also likely only affects a very isolated area near the roots 618 (Dennis et al., 2010), and these inputs cannot easily diffuse in dry soil to be able to influence 619 a large proportion of the soil microbes. An additional consideration is that the length of this 620 experiment may not have been sufficient for demonstrating the full impacts on fungi, which 621 have relatively slow growth compared to bacteria. For example, bacterial turnover time tends 622 to be days to weeks, whereas fungal turnover is usually weeks to months (Rousk and Bååth, 623 2011). The complexity of the microbial community may also mask some impacts that affect 624 some groups of the community. Many specific bacterial and fungal taxa showed an 625 interaction between the drought treatment and the presence of Q. ilex, although the direction 626 of the responses was not consistent. For example, Gammaproteobacteria relative abundance 627 was negatively correlated with increasing drought when Q. ilex was absent, but there was no 26 628 correlation when Q. ilex was present, whereas Cytophagia showed the opposite pattern 629 (negative correlation with drought only when Q. ilex was present). 630 It is important to note that although the roots of the saplings were thoroughly cleaned 631 before planting, it was impossible to completely remove the original community present on 632 the roots. This community was from the substrate that the plants were grown in at the nursery 633 they were purchased from, and could have influenced the microbial community that we 634 measured. However, this should not affect our interpretation of any interaction between the 635 presence of Q. ilex and the drought treatment as the impact of roots being present (with all of 636 the accompanying rhizosphere microbes and root processes) versus roots being absent, would 637 be much greater than any small variations in rhizosphere composition due to the amount of 638 influence of the initial community. Additionally, all roots were treated equally, so there 639 should be no systematic difference in the influence of the initial inoculum for the different 640 plants. 641 642 4.6. Are soil microbial communities with a history of drought more resistant to 643 drought? 644 Recent studies have identified ‘legacy effects’ of drought-stressed soils. This means that for 645 soils that have been previously water stressed, a subsequent drought may have a stronger or 646 weaker impact on the soil community than a soil without such a history (de Vries et al., 2012; 647 Evans and Wallenstein, 2012; Bouskill et al., 2013; Hawkes and Keitt, 2015; Kaisermann et 648 al., 2017; Meisner et al., 2018). A stronger impact could be due to the loss of resistance or 649 resilience of a repeatedly disturbed soil, whereas a weaker impact would imply the selection 650 of taxa that are better adapted to the conditions. We found no interaction between drought 651 and soil history for the overall bacterial or fungal composition or diversity, but some 27 652 evidence suggested that previous soil history affected the response of specific taxa to drought. 653 For example, the relative abundance of the bacterial phylum Chloroflexi was correlated 654 positively with drought duration in the control soil but negatively in the pre-droughted soil. 655 Relative abundance may thus increase during a short-term drought (the current experiment) 656 but not during long-term perturbations (the 16 years of the long-term drought study). 657 Previous studies have found Chloroflexi to increase in relative abundance during drought 658 periods or increasing aridity (Acosta-Martínez et al., 2014; Maestre et al., 2015). However, 659 this is not a universal response, with another recent study finding a decrease in relative 660 abundance under drought (Meisner et al., 2018). 661 Another interesting example is the bacterial class Thermoleophilia, where diversity 662 (number of OTUs) was positively correlated with drought duration, but only in the pre- 663 droughted soil. The abundance of Thermoleophilia was also positively correlated with 664 drought duration (in the control and pre-droughted soils) and was therefore an example of a 665 group of bacteria that may be able to take advantage of recurrent droughts, consistent with 666 previous studies reporting that Thermoleophilia can respond positively to drought (Pereira de 667 Castro et al., 2016; Ochoa-Hueso et al., 2018). Thermoleophilia diversity, however, was not 668 higher in the pre-droughted soil before the start of the current drought experiment, 669 demonstrating the difficulty in both predicting the response of soil microbes and 670 understanding the mechanisms behind any adaptation to drought in long-term droughted soil. 671 Bacterial and fungal alpha diversity before the start of the drought treatment was 672 highest in the pre-droughted soil, which could indicate a source of resilience or resistance for 673 the microbial community, even though the diversity was not significantly higher than in the 674 control soil. Further investigation of this finding may provide evidence of adaptation to water 675 stress. Our results indicate a tendency for plant fungal pathogens to increase under drought, 676 and have a higher abundance in soils with a long-term history of drought. This may be due to 28 677 the damaging impact of drought on plants, which may increase their susceptibility to disease 678 and thereby increase the population of fungal soil pathogens (Brasier, 1996). Pre-droughting 679 had a negative legacy effect on the saprotrophic yeasts, with relative abundance negatively 680 correlated with drought duration in the pre-droughted soil. The abundance of yeasts in this 681 case was not higher in the pre-droughted soil, which may indicate that the long-term 682 historical drought increased the vulnerability of this group to the subsequent drought in this 683 experiment. 684 The soil inoculum used in this study was taken from a Mediterranean holm oak forest, 685 which is exposed to relatively large variation in precipitation within and between years 686 (Ogaya and Peñuelas, 2007; Liu et al., 2015). In the period from 1999-2015, mean annual 687 precipitation was 616.1 mm, and this varied greatly from 379.8 mm (in 2006) to 926.7 mm 688 (in 2010). The majority of rainfall (80%) is in spring (March-May) and autumn (September- 689 November), with less than 10% in summer (June-August). The soil water content varies 690 between ~10% v/v in summer to ~30% v/v in spring and autumn. Whilst the long-term 691 drought treatment did decrease mean soil moisture throughout the study period by 13% 692 compared with control plots (Liu et al., 2015), the high variability in precipitation may reduce 693 the chance of the microbial community demonstrating legacy effects, as the soil can cycle 694 between being very dry and then saturated, and the ‘control’ soil community may already be 695 dominated by phenotypes that can tolerate the dry summer conditions (Curiel Yuste et al., 696 2014). Thus legacy effects may be less pronounced in this Mediterranean system compared 697 with areas that have more uniform precipitation patterns, such as in temperate locations (de 698 Vries et al., 2012; Kaisermann et al., 2017; Meisner et al., 2018), humid continental (Evans 699 and Wallenstein, 2012), or humid tropical sites (Bouskill et al., 2013). 700 29 701 4.7. The future of soil communities under drought 702 Drought is a growing threat around the world, and we have demonstrated complex effects on 703 bacterial and fungal communities that depend on the intensity of the drought, the presence of 704 plants, and previous soil history. We have particularly demonstrated that bacteria may be 705 more negatively affected than fungi in terms of biomass stock and that plants may provide 706 some protection for maintaining microbial diversity, so bare soils may be more at risk. Soils 707 with a history of long-term drought showed a legacy effect, which positively affected the 708 diversity of the bacterial community, presumably due to the adaptation of the soil community 709 to these conditions. We also found, however, many examples of taxa or functional groups 710 with a negative legacy effect due to the historical drought. Also it might indeed be that the 711 negative effect on some taxa represents an alleviation of competition that leaves resources 712 available to other taxa, which in turn can increase the diversity. We therefore cannot assume 713 that soil communities will be able to adapt to the occurrence of more frequent or severe 714 droughts and continue to maintain the same functions. Drought will continue to have impacts 715 on microbial community composition, with a general shift towards an increasing proportion 716 of fungi and a decrease in the mass and diversity of bacteria. 717 718 Acknowledgements 719 Funding was provided by the FP7 S-Clima project PIEF-GA-2013-626234, the European 720 Research Council Synergy grant ERC-2013-726 SyG-610028 IMBALANCE-P, the Spanish 721 Government project CGL2016-79835-P (FERTWARM), the Catalan Government project 722 SGR 2014-274 and the EU ClimMani COST action project (ES1308). We thank J van Hal 723 and J de Gruyter for their assistance with molecular work, and the technicians of the Facultat 30 724 de Ciències at the Autonomous University of Barcelona (UAB) that assisted with the 725 autoclaving of soil. 726 727 Declarations of interest: none 728 729 730 731 References 732 Acosta-Martínez, V., Cotton, J., Gardner, T., Moore-Kucera, J., Zak, J., Wester, D., Cox, S., 733 2014. Predominant bacterial and fungal assemblages in agricultural soils during a record 734 drought/heat wave and linkages to enzyme activities of biogeochemical cycling. Applied Soil 735 Ecology 84, 69-82. 736 Alster, C.J., German, D.P., Lu, Y., Allison, S.D., 2013. Microbial enzymatic responses to 737 drought and to nitrogen addition in a southern California grassland. Soil Biology and 738 Biochemistry 64, 68-79. 739 Bachar, A., Al-Ashhab, A., Soares, M.I., Sklarz, M., Angel, R., Ungar, E., Gillor, O., 2010. 740 Soil microbial abundance and diversity along a low precipitation gradient. Microbial ecology 741 60, 453-461. 742 Balasooriya, W.K., Denef, K., Huygens, D., Boeckx, P., 2014. Translocation and turnover of 743 rhizodeposit carbon within soil microbial communities of an extensive grassland ecosystem. 744 Plant and Soil 376, 61-73. 31 745 Baldrian, P., Merhautová, V., Petránková, M., Cajthaml, T., Šnajdr, J., 2010. Distribution of 746 microbial biomass and activity of extracellular enzymes in a hardwood forest soil reflect soil 747 moisture content. Applied Soil Ecology 46, 177-182. 748 Bapiri, A., Bååth, E., Rousk, J., 2010. Drying–rewetting cycles affect fungal and bacterial 749 growth differently in an arable soil. Microbial ecology 60, 419-428. 750 Bardgett, R.D., Hobbs, P.J., Frostegård, Å., 1996. Changes in soil fungal: bacterial biomass 751 ratios following reductions in the intensity of management of an upland grassland. Biology 752 and Fertility of Soils 22, 261-264. 753 Barnard, R.L., Osborne, C.A., Firestone, M.K., 2015. Changing precipitation pattern alters 754 soil microbial community response to wet-up under a Mediterranean-type climate. ISME J. 755 Bastida, F., Torres, I.F., Andrés‐Abellán, M., Baldrian, P., López‐Mondéjar, R., Větrovský, 756 T., Richnow, H.H., Starke, R., Ondoño, S., García, C., López‐Serrano, F.R., Jehmlich, N., 757 2017. Differential sensitivity of total and active soil microbial communities to drought and 758 forest management. Global Change Biology 23, 4185-4203. 759 Beier, C., Beierkuhnlein, C., Wohlgemuth, T., Penuelas, J., Emmett, B., Koerner, C., de 760 Boeck, H.J., Christensen, J.H., Leuzinger, S., Janssens, I.A., Hansen, K., 2012. Precipitation 761 manipulation experiments - challenges and recommendations for the future. Ecology Letters 762 15, 899-911. 763 Bell, C., McIntyre, N., Cox, S., Tissue, D., Zak, J., 2008. Soil microbial responses to 764 temporal variations of moisture and temperature in a Chihuahuan desert grassland. Microbial 765 ecology 56, 153-167. 766 Bergsma-Vlami, M., Prins, M.E., Raaijmakers, J.M., 2005. Influence of plant species on 767 population dynamics, genotypic diversity and antibiotic production in the rhizosphere by 768 indigenous Pseudomonas spp. FEMS Microbiology Ecology 52, 59-69. 32 769 Borken, W., Matzner, E., 2009. Reappraisal of drying and wetting effects on C and N 770 mineralization and fluxes in soils. Global Change Biology 15, 808-824. 771 Bouskill, N.J., Lim, H.C., Borglin, S., Salve, R., Wood, T.E., Silver, W.L., Brodie, E.L., 772 2013. Pre-exposure to drought increases the resistance of tropical forest soil bacterial 773 communities to extended drought. Isme Journal 7, 384-394. 774 Brasier, C.M., 1996. Phytophthora cinnamomi and oak decline in southern Europe. 775 Environmental constraints including climate change. Annales Des Sciences Forestieres 53, 776 347-358. 777 Bressan, M., Roncato, M.-A., Bellvert, F., Comte, G., Haichar, F.e.Z., Achouak, W., Berge, 778 O., 2009. Exogenous glucosinolate produced by Arabidopsis thaliana has an impact on 779 microbes in the rhizosphere and plant roots. The Isme Journal 3, 1243. 780 Caporaso, J.G., Bittinger, K., Bushman, F.D., DeSantis, T.Z., Andersen, G.L., Knight, R., 781 2010a. PyNAST: a flexible tool for aligning sequences to a template alignment. 782 Bioinformatics 26, 266-267. 783 Caporaso, J.G., Kuczynski, J., Stombaugh, J., Bittinger, K., Bushman, F.D., Costello, E.K., 784 Fierer, N., Pena, A.G., Goodrich, J.K., Gordon, J.I., 2010b. QIIME allows analysis of high- 785 throughput community sequencing data. Nature methods 7, 335. 786 Castaño, C., Lindahl, B.D., Alday, J.G., Hagenbo, A., Martínez de Aragón, J., Parladé, J., 787 Pera, J., Bonet, J.A., 2018. Soil microclimate changes affect soil fungal communities in a 788 Mediterranean pine forest. New Phytologist. 789 Chowdhury, N., Marschner, P., Burns, R., 2011. Response of microbial activity and 790 community structure to decreasing soil osmotic and matric potential. Plant and Soil 344, 241- 791 254. 792 Cook, R., Papendick, R., 1972. Influence of water potential of soils and plants on root 793 disease. Annual Review of Phytopathology 10, 349-374. 33 794 Cregger, M.A., Schadt, C.W., McDowell, N.G., Pockman, W.T., Classen, A.T., 2012. 795 Response of the soil microbial community to changes in precipitation in a semiarid 796 ecosystem. Appl Environ Microbiol 78, 8587-8594. 797 Crowther, T.W., Sokol, N.W., Oldfield, E.E., Maynard, D.S., Thomas, S.M., Bradford, M.A., 798 2015. Environmental stress response limits microbial necromass contributions to soil organic 799 carbon. Soil Biology and Biochemistry 85, 153-161. 800 Curiel Yuste, J., Fernandez-Gonzalez, A.J., Fernandez-Lopez, M., Ogaya, R., Penuelas, J., 801 Sardans, J., Lloret, F., 2014. Strong functional stability of soil microbial communities under 802 semiarid Mediterranean conditions and subjected to long-term shifts in baseline precipitation. 803 Soil Biology & Biochemistry 69, 223-233. 804 de Sampaio e Paiva Camilo-Alves, C., da Clara, M.I.E., de Almeida Ribeiro, N.M.C., 2013. 805 Decline of Mediterranean oak trees and its association with Phytophthora cinnamomi: a 806 review. European Journal of Forest Research 132, 411-432. 807 de Vries, F.T., Griffiths, R.I., Bailey, M., Craig, H., Girlanda, M., Gweon, H.S., Hallin, S., 808 Kaisermann, A., Keith, A.M., Kretzschmar, M., 2018. Soil bacterial networks are less stable 809 under drought than fungal networks. Nature Communications 9, 3033. 810 de Vries, F.T., Liiri, M.E., Bjornlund, L., Setala, H.M., Christensen, S., Bardgett, R.D., 2012. 811 Legacy effects of drought on plant growth and the soil food web. Oecologia 170, 821-833. 812 Dennis, P.G., Miller, A.J., Hirsch, P.R., 2010. Are root exudates more important than other 813 sources of rhizodeposits in structuring rhizosphere bacterial communities? FEMS 814 Microbiology Ecology 72, 313-327. 815 DeSantis, T.Z., Hugenholtz, P., Larsen, N., Rojas, M., Brodie, E.L., Keller, K., Huber, T., 816 Dalevi, D., Hu, P., Andersen, G.L., 2006. Greengenes, a chimera-checked 16S rRNA gene 817 database and workbench compatible with ARB. Appl Environ Microbiol 72, 5069-5072. 34 818 Desprez-Loustau, M.-L., Marçais, B., Nageleisen, L.-M., Piou, D., Vannini, A., 2006. 819 Interactive effects of drought and pathogens in forest trees. Annals of Forest Science 63, 597- 820 612. 821 Edgar, R.C., 2013. UPARSE: highly accurate OTU sequences from microbial amplicon 822 reads. Nature methods 10, 996. 823 Evans, S.E., Wallenstein, M.D., 2012. Soil microbial community response to drying and 824 rewetting stress: does historical precipitation regime matter? Biogeochemistry 109, 101-116. 825 Field, C.B., Barros, V.R., Mach, K.J., Mastrandrea, M.D., Aalst, M.v., Adger, W.N., Arent, 826 D.J., Barnett, J., Betts, R., Bilir, T.E., Birkmann, J., Carmin, J., Chadee, D.D., Challinor, A.J., 827 Chatterjee, M., Cramer, W., Davidson, D.J., Estrada, Y.O., Gattuso, J.P., Hijioka, Y., Hoegh- 828 Guldberg, O., Huang, H.Q., Insarov, G.E., Jones, R.N., Kovats, R.S., Lankao, P.R., Larsen, 829 J.N., Losada, I.J., Marengo, J.A., McLean, R.F., Mearns, L.O., Mechler, R., Morton, J.F., 830 Niang, I., Oki, T., Olwoch, J.M., Opondo, M., Poloczanska, E.S., Pörtner, H.O., Redsteer, 831 M.H., Reisinger, A., Revi, A., Schmidt, D.N., Shaw, M.R., Solecki, W., Stone, D.A., Stone, 832 J.M.R., Strzepek, K.M., Suarez, A.G., Tschakert, P., Valentini, R., Vicuña, S., Villamizar, A., 833 Vincent, K.E., Warren, R., White, L.L., Wilbanks, T.J., Wong, P.P., Yohe, G.W., 2014. 834 Technical Summary, In: Field, C.B., Barros, V.R., Dokken, D.J., Mach, K.J., Mastrandrea, 835 M.D., Bilir, T.E., Chatterjee, M., Ebi, K.L., Estrada, Y.O., Genova, R.C., Girma, B., Kissel, 836 E.S., Levy, A.N., MacCracken, S., Mastrandrea, P.R., White, L.L. (Eds.), Climate Change 837 2014: Impacts, Adaptation, and Vulnerability. Part A: Global and Sectoral Aspects. 838 Contribution of Working Group II to the Fifth Assessment Report of the Intergovernmental 839 Panel on Climate Change. Cambridge University Press, Cambridge, United Kingdom and 840 New York, NY, USA, pp. 35-94. 841 Frank, D., Reichstein, M., Bahn, M., Thonicke, K., Frank, D., Mahecha, M.D., Smith, P., van 842 der Velde, M., Vicca, S., Babst, F., Beer, C., Buchmann, N., Canadell, J.G., Ciais, P., 35 843 Cramer, W., Ibrom, A., Miglietta, F., Poulter, B., Rammig, A., Seneviratne, S.I., Walz, A., 844 Wattenbach, M., Zavala, M.A., Zscheischler, J., 2015. Effects of climate extremes on the 845 terrestrial carbon cycle: concepts, processes and potential future impacts. Global Change 846 Biology 21, 2861–2880. 847 Frostegård, Å., Bååth, E., 1996. The use of phospholipid fatty acid analysis to estimate 848 bacterial and fungal biomass in soil. Biology and Fertility of Soils 22, 59-65. 849 Frostegård, Å., Bååth, E., Tunlio, A., 1993. Shifts in the structure of soil microbial 850 communities in limed forests as revealed by phospholipid fatty acid analysis. Soil Biology 851 and Biochemistry 25, 723-730. 852 Frostegård, Å., Tunlid, A., Bååth, E., 2011. Use and misuse of PLFA measurements in soils. 853 Soil Biology and Biochemistry 43, 1621-1625. 854 Fuchslueger, L., Bahn, M., Fritz, K., Hasibeder, R., Richter, A., 2014. Experimental drought 855 reduces the transfer of recently fixed plant carbon to soil microbes and alters the bacterial 856 community composition in a mountain meadow. New Phytologist 201, 916-927. 857 Griffiths, B.S., Philippot, L., 2013. Insights into the resistance and resilience of the soil 858 microbial community. FEMS microbiology reviews 37, 112-129. 859 Haichar, F.e.Z., Marol, C., Berge, O., Rangel-Castro, J.I., Prosser, J.I., Balesdent, J., Heulin, 860 T., Achouak, W., 2008. Plant host habitat and root exudates shape soil bacterial community 861 structure. ISME J 2, 1221-1230. 862 Harris, J., 2003. Measurements of the soil microbial community for estimating the success of 863 restoration. European Journal of Soil Science 54, 801-808. 864 Hartmann, M., Brunner, I., Hagedorn, F., Bardgett, R.D., Stierli, B., Herzog, C., Chen, X., 865 Zingg, A., Graf-Pannatier, E., Rigling, A., Frey, B., 2017. A decade of irrigation transforms 866 the soil microbiome of a semi-arid pine forest. Molecular Ecology 26, 1190-1206. 36 867 Hawkes, C.V., Keitt, T.H., 2015. Resilience vs. historical contingency in microbial responses 868 to environmental change. Ecology Letters 18, 612-625. 869 Holland, E., Coleman, D.C., 1987. Litter placement effects on microbial and organic matter 870 dynamics in an agroecosystem. Ecology 68, 425-433. 871 Hoover, D.L., Rogers, B.M., 2016. Not all droughts are created equal: the impacts of 872 interannual drought pattern and magnitude on grassland carbon cycling. Global Change 873 Biology 22, 1809-1820. 874 Hortal, S., Bastida, F., Moreno, J.L., Armas, C., García, C., Pugnaire, F.I., 2015. Benefactor 875 and allelopathic shrub species have different effects on the soil microbial community along 876 an environmental severity gradient. Soil Biology and Biochemistry 88, 48-57. 877 Hueso, S., Garcia, C., Hernandez, T., 2012. Severe drought conditions modify the microbial 878 community structure, size and activity in amended and unamended soils. Soil Biology & 879 Biochemistry 50, 167-173. 880 Kaisermann, A., de Vries, F.T., Griffiths, R.I., Bardgett, R.D., 2017. Legacy effects of 881 drought on plant–soil feedbacks and plant–plant interactions. New Phytologist 215, 1413- 882 1424. 883 Kaisermann, A., Maron, P.A., Beaumelle, L., Lata, J.C., 2015. Fungal communities are more 884 sensitive indicators to non-extreme soil moisture variations than bacterial communities. 885 Applied Soil Ecology 86, 158-164. 886 Klindworth, A., Pruesse, E., Schweer, T., Peplies, J., Quast, C., Horn, M., Glöckner, F.O., 887 2013. Evaluation of general 16S ribosomal RNA gene PCR primers for classical and next- 888 generation sequencing-based diversity studies. Nucleic acids research 41, e1-e1. 889 Knapp, A.K., Ciais, P., Smith, M.D., 2017. Reconciling inconsistencies in precipitation– 890 productivity relationships: implications for climate change. New Phytologist 214, 41-47. 37 891 Kõljalg, U., Larsson, K.H., Abarenkov, K., Nilsson, R.H., Alexander, I.J., Eberhardt, U., 892 Erland, S., Høiland, K., Kjøller, R., Larsson, E., 2005. UNITE: a database providing web‐ 893 based methods for the molecular identification of ectomycorrhizal fungi. New Phytologist 894 166, 1063-1068. 895 Ladygina, N., Hedlund, K., 2010. Plant species influence microbial diversity and carbon 896 allocation in the rhizosphere. Soil Biology and Biochemistry 42, 162-168. 897 Lange, M., Eisenhauer, N., Sierra, C.A., Bessler, H., Engels, C., Griffiths, R.I., Mellado- 898 Vázquez, P.G., Malik, A.A., Roy, J., Scheu, S., 2015. Plant diversity increases soil microbial 899 activity and soil carbon storage. Nature Communications 6, 6707. 900 Lareen, A., Burton, F., Schäfer, P., 2016. Plant root-microbe communication in shaping root 901 microbiomes. Plant Molecular Biology 90, 575-587. 902 Larsen, K.S., Andresen, L.C., Beier, C., Jonasson, S., Albert, K.R., Ambus, P., Arndal, M.F., 903 Carter, M.S., Christensen, S., Holmstrup, M., 2011. Reduced N cycling in response to 904 elevated CO2, warming, and drought in a Danish heathland: synthesizing results of the 905 CLIMAITE project after two years of treatments. Global Change Biology 17, 1884-1899. 906 Liu, D., Ogaya, R., Barbeta, A., Yang, X., Peñuelas, J., 2015. Contrasting impacts of 907 continuous moderate drought and episodic severe droughts on the aboveground-biomass 908 increment and litterfall of three coexisting Mediterranean woody species. Global Change 909 Biology 21, 4196-4209. 910 Maestre, F.T., Delgado-Baquerizo, M., Jeffries, T.C., Eldridge, D.J., Ochoa, V., Gozalo, B., 911 Quero, J.L., García-Gómez, M., Gallardo, A., Ulrich, W., Bowker, M.A., Arredondo, T., 912 Barraza-Zepeda, C., Bran, D., Florentino, A., Gaitán, J., Gutiérrez, J.R., Huber-Sannwald, E., 913 Jankju, M., Mau, R.L., Miriti, M., Naseri, K., Ospina, A., Stavi, I., Wang, D., Woods, N.N., 914 Yuan, X., Zaady, E., Singh, B.K., 2015. Increasing aridity reduces soil microbial diversity 38 915 and abundance in global drylands. Proceedings of the National Academy of Sciences 112, 916 15684-15689. 917 Manzoni, S., Schimel, J.P., Porporato, A., 2011. Responses of soil microbial communities to 918 water stress: results from a meta-analysis. Ecology 93, 930-938. 919 Martinez Arbizu, P., 2017. pairwiseAdonis: Pairwise multilevel comparison using adonis. R 920 package version 0.0.1. 921 Meisner, A., Jacquiod, S., Snoek, B.L., ten Hooven, F.C., van der Putten, W.H., 2018. 922 Drought legacy effects on the composition of soil fungal and prokaryote communities. 923 Frontiers in Microbiology 9. 924 Mengtian, H., Xuhui, W., F., K.T., Shilong, P., 2018. Drought timing influences the legacy of 925 tree growth recovery. Global Change Biology 24, 3546-3559. 926 Moore, J.C., Berlow, E.L., Coleman, D.C., Ruiter, P.C., Dong, Q., Hastings, A., Johnson, 927 N.C., McCann, K.S., Melville, K., Morin, P.J., Nadelhoffer, K., Rosemond, A.D., Post, D.M., 928 Sabo, J.L., Scow, K.M., Vanni, M.J., Wall, D.H., 2004. Detritus, trophic dynamics and 929 biodiversity. Ecology Letters 7, 584-600. 930 Naylor, D., Coleman-Derr, D., 2017. Drought stress and root-associated bacterial 931 communities. Frontiers in Plant Science 8. 932 Nessner Kavamura, V., Taketani, R.G., Lançoni, M.D., Andreote, F.D., Mendes, R., Soares 933 de Melo, I., 2013. Water regime influences bulk soil and rhizosphere of Cereus jamacaru 934 bacterial communities in the Brazilian Caatinga biome. Plos One 8, e73606. 935 Nguyen, L.T.T., Osanai, Y., Anderson, I.C., Bange, M.P., Tissue, D.T., Singh, B.K., 2018. 936 Flooding and prolonged drought have differential legacy impacts on soil nitrogen cycling, 937 microbial communities and plant productivity. Plant and Soil 431, 371-387. 938 Ochoa-Hueso, R., Collins, S.L., Delgado‐Baquerizo, M., Hamonts, K., Pockman, W.T., 939 Sinsabaugh, R.L., Smith, M.D., Knapp, A.K., Power, S.A., 2018. Drought consistently alters 39 940 the composition of soil fungal and bacterial communities in grasslands from two continents. 941 Global Change Biology. 942 Ogaya, R., Peñuelas, J., 2007. Tree growth, mortality, and above-ground biomass 943 accumulation in a holm oak forest under a five-year experimental field drought. Plant 944 Ecology 189, 291-299. 945 Pereira de Castro, A., Sartori da Silva, M.R.S., Quirino, B.F., da Cunha Bustamante, M.M., 946 Krüger, R.H., 2016. Microbial diversity in Cerrado biome (neotropical savanna) soils. Plos 947 One 11, e0148785. 948 Philippot, L., Raaijmakers, J.M., Lemanceau, P., van der Putten, W.H., 2013. Going back to 949 the roots: the microbial ecology of the rhizosphere. Nat Rev Micro 11, 789-799. 950 Preece, C., Farré-Armengol, G., Llusià, J., Peñuelas, J., 2018. Thirsty tree roots exude more 951 carbon. Tree Physiology 38, 690–695. 952 Preece, C., Peñuelas, J., 2016. Rhizodeposition under drought and consequences for soil 953 communities and ecosystem resilience. Plant and Soil 409, 1-17. 954 R Core Team, 2016. R: A language and environment for statistical computing, R Foundation 955 for Statistical Computing, Vienna, Austria. . 956 Rousk, J., Bååth, E., 2011. Growth of saprotrophic fungi and bacteria in soil. FEMS 957 Microbiology Ecology 78, 17-30. 958 Sanaullah, M., Blagodatskaya, E., Chabbi, A., Rumpel, C., Kuzyakov, Y., 2011. Drought 959 effects on microbial biomass and enzyme activities in the rhizosphere of grasses depend on 960 plant community composition. Applied Soil Ecology 48, 38-44. 961 Sanaullah, M., Rumpel, C., Charrier, X., Chabbi, A., 2012. How does drought stress 962 influence the decomposition of plant litter with contrasting quality in a grassland ecosystem? 963 Plant and Soil 352, 277-288. 40 964 Schimel, J., Balser, T.C., Wallenstein, M., 2007. Microbial stress-response physiology and its 965 implications for ecosystem function. Ecology 88, 1386-1394. 966 Schmidt, P.A., Schmitt, I., Otte, J., Bandow, C., Rombke, J., Balint, M., Rolshausen, G., 967 2018. Season-long experimental drought alters fungal community composition but not 968 diversity in a grassland soil. Microbial ecology 75, 468-478. 969 Smith, D.P., Peay, K.G., 2014. Sequence depth, not PCR replication, improves ecological 970 inference from next generation DNA sequencing. Plos One 9, e90234. 971 Strickland, M.S., Rousk, J., 2010. Considering fungal:bacterial dominance in soils – 972 Methods, controls, and ecosystem implications. Soil Biology and Biochemistry 42, 1385- 973 1395. 974 Taketani, R.G., Lançoni, M.D., Kavamura, V.N., Durrer, A., Andreote, F.D., Melo, I.S., 975 2017. Dry season constrains bacterial phylogenetic diversity in a semi-arid rhizosphere 976 system. Microbial ecology 73, 153-161. 977 Tedersoo, L., Bahram, M., Põlme, S., Kõljalg, U., Yorou, N.S., Wijesundera, R., Ruiz, L.V., 978 Vasco-Palacios, A.M., Thu, P.Q., Suija, A., 2014. Global diversity and geography of soil 979 fungi. Science 346, 1256688. 980 Thakur, M.P., Milcu, A., Manning, P., Niklaus, P.A., Roscher, C., Power, S., Reich, P.B., 981 Scheu, S., Tilman, D., Ai, F.X., Guo, H.Y., Ji, R., Pierce, S., Ramirez, N.G., Richter, A.N., 982 Steinauer, K., Strecker, T., Vogel, A., Eisenhauer, N., 2015. Plant diversity drives soil 983 microbial biomass carbon in grasslands irrespective of global environmental change factors. 984 Global Change Biology 21, 4076-4085. 985 Tóth, Z., Táncsics, A., Kriszt, B., Kröel‐Dulay, G., Ónodi, G., Hornung, E., 2017. Extreme 986 effects of drought on composition of the soil bacterial community and decomposition of plant 987 tissue. European Journal of Soil Science 68, 504-513. 41 988 Touma, D., Ashfaq, M., Nayak, M.A., Kao, S.-C., Diffenbaugh, N.S., 2015. A multi-model 989 and multi-index evaluation of drought characteristics in the 21st century. Journal of 990 Hydrology 526, 196-207. 991 Treseder, K.K., Lennon, J.T., 2015. Fungal traits that drive ecosystem dynamics on land. 992 Microbiology and Molecular Biology Reviews 79, 243-262. 993 Tunlid, A., Hoitink, H., Low, C., White, D., 1989. Characterization of bacteria that suppress 994 Rhizoctonia damping-off in bark compost media by analysis of fatty acid biomarkers. Appl 995 Environ Microbiol 55, 1368-1374. 996 United Nations Convention to Combat Desertification, 2017. The Global Land Outlook, first 997 edition. Bonn, Germany. 998 van der Molen, M.K., Dolman, A.J., Ciais, P., Eglin, T., Gobron, N., Law, B.E., Meir, P., 999 Peters, W., Phillips, O.L., Reichstein, M., Chen, T., Dekker, S.C., Doubková, M., Friedl, 1000 M.A., Jung, M., van den Hurk, B.J.J.M., de Jeu, R.A.M., Kruijt, B., Ohta, T., Rebel, K.T., 1001 Plummer, S., Seneviratne, S.I., Sitch, S., Teuling, A.J., van der Werf, G.R., Wang, G., 2011. 1002 Drought and ecosystem carbon cycling. Agricultural and Forest Meteorology 151, 765-773. 1003 Wang, Q., Garrity, G.M., Tiedje, J.M., Cole, J.R., 2007. Naïve Bayesian classifier for rapid 1004 assignment of rRNA sequences into the new bacterial taxonomy. Appl Environ Microbiol 73, 1005 5261-5267. 1006 Weedon, J.T., Aerts, R., Kowalchuk, G.A., van Logtestijn, R., Andringa, D., van Bodegom, 1007 P.M., 2013. Temperature sensitivity of peatland C and N cycling: Does substrate supply play 1008 a role? Soil Biology and Biochemistry 61, 109-120. 1009 Yin, B., Crowley, D., Sparovek, G., De Melo, W.J., Borneman, J., 2000. Bacterial functional 1010 redundancy along a soil reclamation gradient. Appl Environ Microbiol 66, 4361-4365. 42 1011 Yuste, J.C., Peñuelas, J., Estiarte, M., Garcia‐Mas, J., Mattana, S., Ogaya, R., Pujol, M., 1012 Sardans, J., 2011. Drought‐resistant fungi control soil organic matter decomposition and its 1013 response to temperature. Global Change Biology 17, 1475-1486. 1014 Zak, D.R., Holmes, W.E., White, D.C., Peacock, A.D., Tilman, D., 2003. Plant diversity, soil 1015 microbial communities, and ecosystem function: are there any links? Ecology 84, 2042-2050. 1016 Zelles, L., 1999. Fatty acid patterns of phospholipids and lipopolysaccharides in the 1017 characterisation of microbial communities in soil: a review. Biology and Fertility of Soils 29, 1018 111-129. 1019 1020 1021 1022 1023 1024 43 1025 1026 Bacteria Fungi Treatment Composition Shannon Simpson Composition Shannon Simpson factor diversity index diversity index R2 P χ² P χ² P R2 P χ² P χ² P Soil history 0.25 *** NS NS 6.3 * 0.51 *** 11.9 ** NS NS Plant 0.16 *** NS NS NS NS NS NS NS NS NS NS Soil history NS NS 29.5 *** 21.9 *** NS NS NS NS 9.3 ** * Plant 1027 Table 1. Factors affecting initial differences in soil community composition (results of PERMANOVA using the adonis2 function in the R vegan 1028 package) and Shannon diversity (results of linear mixed effects model). Asterisks represent the P-value:* = P < 0.05, ** = P < 0.01, *** = P < 0.001. 1029 NS = non-significant. 1030 1031 44 Treatment factor Bacteria Fungi Type of drought Continuous Factor Continuous Factor variable R2 P R2 P R2 P R2 P Drought 0.12 *** 0.06 *** NS NS NS NS Soil history 0.17 *** 0.17 *** 0.40 *** 0.40 *** Plant 0.08 *** 0.08 *** 0.04 *** 0.04 *** Drought * Soil NS NS NS NS NS NS NS NS Drought * Plant NS NS 0.03 * NS NS NS NS Soil history * Plant 0.03 *** 0.03 *** 0.03 * 0.03 * Table 2. Results of the PERMANOVA of microbial composition (Bray-Curtis dissimilarity indices) using the adonis2 function in the R vegan package. Drought was treated as both a continuous variable (days of drought) and as a factor (control, low, mid, high), as the impact of drought may not always be linear. Asterisks represent the P-value: * = P < 0.05 and *** = P < 0.001. Bacteria Fungi Shannon Simpson Shannon Simpson Factor χ² P χ² P χ² P χ² P Drought 9.0 ** 6.2 * 21.6 *** 19.0 *** Soil history 12.7 ** NS NS 15.4 *** NS NS Plant NS NS NS NS NS NS NS NS Soil history * plant 37.6 *** 19.4 *** 9.0 * 13.6 ** Table 3. Effect of drought, soil history, and the presence of Q. ilex on the Shannon and Simpson index of bacteria and fungi. Drought duration is a continuous variable for bacteria and a categorical variable for fungi as the response to drought was non-linear. 45 Soil history Q. ilex H Bacteria Fungi Control Present -2.91 +26.04 Control Absent -1.71 +8.33 Pre-droughted Present -1.58 -1.46 Pre-droughted Absent -1.84 +17.47 Pre-sterilised Present -4.08 +15.15 Pre-sterilised Absent -3.94 +34.65 Table 4. H (% change after most intense drought compared to control) of the Shannon diversity (H) of the bacterial and fungal communities to drought. 46 Fig. 1 (a) Plant effect (b) Soil history effect 1.0 1.0 control With Q . i l ex pre−droughted Soil only pre−sterilised 0.5 0.5 Bacteria NMDS2 NMDS2 0.0 0.0 −0.5 −0.5 −1.0 −1.0 −0.5 0.0 0.5 1.0 −0.5 0.0 0.5 1.0 NMDS1 NMDS1 (c) (d) 1.5 1.5 1.0 1.0 Fungi NMDS2 NMDS2 0.5 0.5 0.0 0.0 −0.5 −0.5 −1.0 −1.0 −1.5 −0.5 0.5 1.0 −1.5 −0.5 0.5 1.0 NMDS1 NMDS1 Fig. 1. NMDS plots based on Bray-Curtis dissimilarities of the (a,b) bacterial and (c,d) fungal communities prior to onset of drought (drought duration of zero days), showing the difference due to the presence of Q. ilex and the soil history. Points represent individual samples, convex hulls encompass samples of the same treatment. 47 Fig. 2 With Q. ilex Soil only (a) With Q. ilex Soil only (b) 70 ● 0.125 ● 60 Total PLFAs (nmol. g-1 ) ● ● Fungal:bacterial ratio ● 0.100 ● ● 50 ● ● ● ● ● ● ● ● ● ● 0.075 ● ● ● ●● ●● 40 ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● 30 ● 0.050 ●● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ●● ● ● ● ● ● ● ● 20 0 5 10 15 20 Control Drought Sterile Control Drought Sterile Days of drought Soil history Fig. 2. (a) Fungal:bacterial ratio for the drought durations with Q. ilex present (green lines and dots) and for soil only (black lines and grey dots) There was a positive relationship with drought (P < 0.01) and a positive effect of plant-presence (P < 0.001). (b) Total PLFAs (nmol g-1) separated by soil history and presence of Q. ilex (blue) or soil only (red). n = 15 for each plant and soil combination. Soil treatments are ‘Control’, ‘Drought’ (pre-droughted) and ‘Sterile’ (pre-sterilised). 48 Fig. 3. NMDS plots based on Bray-Curtis dissimilarities of the (a) bacterial and (b) fungal communities after drought treatments of 0 to 21 days. Numbers (0 to 21) show the length in days of the drought treatment, and represent individual samples. The six combinations of soil history and plant presence are grouped by colour, with shades of red when Q. ilex was present and shades of blue when Q. ilex was absent. Convex hulls encompass samples of the same treatment. 49 Fig. 4 (a) (b) With Q. ilex Soil only 6.4 ● ● 6.4 ● ● ● ● ● ● ● ● ● ● ●● ● ● Shannon diversity (H) ● ●● ● ● ● ● ● ● ● ● 6.0 ● ● ● ● ● ● ● ● ● 6.0 ●● ●● ● ● ● ● ● ● ● ● ● ● ●●●● ● ● ● ●● ● ● ● ● ● ●● ● ● ● ●● ● ● ● ● 5.6 ● ● 5.6 ● ● ● ● ● ●● ● ● ● ● 5.2 ● ● 5.2 ● ● ● 4.8 4.8 0 5 10 15 20 Control Drought Sterile Control Drought Sterile Days of drought Soil history Fig. 4. Relationships between bacterial Shannon diversity and (a) drought duration (days) (P < 0.001) and (b) soil history (P < 0.01, all three soil histories differed from each other) and the presence of Q. ilex (significant interaction, P < 0.001). n = 15 for each plant and soil combination. Soil treatments are ‘Control’, ‘Drought’ (pre-droughted) and ‘Sterile’ (pre- sterilised). For plot (a) small grey points represent individual samples and larger black points are mean values. 50 Fig. 5 (a) (b) With Q. ilex Soil only 4 a b ab 4 b Shannon diversity (H) 3 3 2 2 1 1 0 0 Control Low Mid High Control Drought Sterile Control Drought Sterile Drought intensity Soil history Fig. 5. Relationships between fungal Shannon diversity and (a) drought duration as a categorical variable (Control, 0 days of drought; Low, 2-7 days; Mid, 9-14 days; High, 16-21 days; P < 0.001) and (b) soil history (P < 0.01, pre-sterilised soil differs from the control and pre-droughted soil) and the presence of Q. ilex (significant interaction, P < 0.001). Letters a and b denote treatments which differ from each other (Tukey HSD test, P < 0.05). n = 15 for each plant and soil combination except those with Q. ilex and in pre-droughted soil, where n = 14. Soil treatments are ‘Control’, ‘Drought’ (pre-droughted) and ‘Sterile’ (pre-sterilised). 51 Supporting Information for: Effects of past and current drought on the composition and diversity of soil microbial communities Catherine Preece, Erik Verbruggen, Lei Liu, James Weedon, Josep Peñuelas. Drought Soil only With Plant duration (d) Control Drought Sterile Control Drought Sterile n n n n n n 0 2 3 3 2 3 3 2 1 1 1 1 1 1 4 1 1 1 1 1 1 7 1 1 1 1 1 1 9 3 2 3 3 3 2 11 1 1 1 1 1 1 14 1 1 1 1 1* 1 16 1 1 1 1 1 1 18 1 1 1 1 1** 1 21 3 3 2 3 2 3 Table S1. The number of replicates shown for each combination of drought, soil and plant treatment. * For the fungi, this sample was not able to be used, thus there was no replicate for the fungal analyses. ** This sample was not able to be used for either bacteria or fungi. Treatment factor Bacteria Fungi Type of drought variable Continuous Factor Continuous Factor 2 2 2 2 R P R P R P R P Drought 0.19 *** 0.09 *** NS NS 0.07 * Soil history 0.05 *** 0.05 *** 0.09 *** 0.10 *** Plant 0.09 *** 0.09 *** 0.10 *** 0.10 *** Drought * Soil history NS NS NS NS NS NS NS NS Drought * Plant NS NS 0.05 * NS NS NS NS Soil history * Plant 0.02 ** 0.02 * NS NS NS NS Table S2. Results of the PERMANOVA of microbial composition (Bray-Curtis dissimilarity indices) when the pre-sterilised soil was not included. The adonis2 function in the R vegan Package was used. Drought was treated as both a continuous variable (days of drought) and as a factor (control, low, mid, high), as the impact of drought may not always be linear. Asterisks represent the P-value: * = P < 0.05, ** = P < 0.01, and *** = P < 0.001. 52 Pairs n F R2 P adj. Bacteria Dcon – Dlow 1.65 0.049 0.198 Dcon – Dmid 2.14 0.049 0.018 Dcon – Dhigh 2.55 0.059 0.006 Dlow – Dmid 1.00 0.022 1.000 Dlow – Dhigh 1.49 0.033 0.366 Dmid – Dhigh 2.04 0.037 0.060 Fungi Dcon –Dlow 0.86 0.028 1.000 Dcon – Dmid 0.69 0.017 1.000 Dcon – Dhigh 0.93 0.022 1.000 Dlow – Dmid 0.96 0.023 1.000 Dlow – Dhigh 0.82 0.019 1.000 Dmid – Dhigh 0.93 0.018 1.000 Table S3. Results of the pairwise PERMANOVA plots of microbial composition (Bray-Curtis dissimilarity indices) between different drought levels (with drought as a factor). Drought levels are: Dcon = control (0 days drought), Dlow = low (2-7 days drought), Dmid = mid (9-14 days drought), Dhigh = high (16-21 days drought). Pairs n F R2 P adj. Bacteria Scon – Sdro 2.93 0.049 0.003 Scon – Sste 11.76 0.169 0.003 Sdro– Sste 12.16 0.176 0.003 Fungi Scon – Sdro 5.95 0.096 0.003 Scon – Sste 33.65 0.384 0.003 Sdro– Sste 40.76 0.421 0.003 Table S4. Results of pairwise PERMANOVA of microbial composition (Bray-Curtis dissimilarity indices) between different soil histories. Soil histories are: Scon = control, Sdro = pre-droughted, Sste = pre-sterilised. 53 Phylum Change in relative abundance Drought * Soil Drought * Plant Plant * Soil Drought effect Soil history Plant effect history history effect Proteobacteria NS NS NS -ve correlation pre-sterilised soil only < soil > control = with Q. ilex drought Actinobacteria NS NS NS +ve correlation pre-sterilised soil only > soil < control = with Q. ilex drought Bacteroidetes -ve correlation in NS soil only in pre- NS NS NS control soil and no sterilised soil > correlation in pre- other combinations sterilised soil Verrucomicrobia NS NS NS pre-sterilised > soil only < control with Q. ilex. Planctomycetes NS NS in soil only, pre- +ve correlation NS NS sterilised soil < control = drought, and soil only with pre-sterilised soil < with Q. ilex in drought soil Chloroflexi +ve correlation in NS in soil only, pre- NS NS NS control soil, -ve sterilised < control correlation in pre- = drought; with Q. droughted soil ilex, pre-sterilised = control < drought Acidobacteria NS NS soil only in pre- NS NS NS sterilised soil > other combinations 54 Firmicutes NS +ve correlation NS NS pre-sterilised > NS when Q. ilex control = present, -ve drought correlation in soil only Table S5. Summary of differences in relative abundances of bacterial phyla contributing > 2% of total reads during the drought experiment. NS, not significant. +ve = positive and =ve = negative. 55 Class Change in relative abundance Drought * Soil Drought * Plant Plant * Soil history Drought effect Soil history Plant effect history effect Alphaproteobacteria NS NS NS NS NS NS Actinobacteria NS NS NS +ve correlation NS soil only > with Q. ilex Betaproteobacteria NS NS NS NS pre-sterilised > NS control = drought Gammaproteobacteria NS -ve correlation NS NS NS NS when soil only, no correlation when Q. ilex present Saprospirae NS NS soil only in pre- -ve correlation NS NS sterilised soil > other combinations Thermoleophilia no correlation in NS NS NS NS NS pre-sterilised soil, +ve correlation in control and drought soils Cytophagia +ve correlation in no correlation in NS NS NS NS pre-sterilised soil only, -ve soils, -ve correlation when correlation in Q. ilex present control soils, no effect in drought soil Deltaproteobacteria NS NS +ve effect of Q. ilex NS NS NS in pre-sterilised soil 56 Planctomycetia NS NS NS +ve correlation NS NS Opitutae NS NS +ve effect of Q. ilex -ve correlation NS NS in control soil Bacilli NS no correlation in NS NS pre-sterilised NS soil only, +ve soil > control = correlation when drought Q. ilex present Anaerolineae NS NS +ve effect of Q. ilex NS NS NS in pre-sterilised and drought Table S6. Summary of differences in relative abundances of bacterial classes contributing > 2% of total reads during the drought experiment. NS, not significant. +ve = positive and =ve = negative. 57 Functional group Change in relative abundance Drought * Soil Drought * Plant Plant * Soil Drought effect Soil history Plant effect history history effect Filamentous NS NS NS NS pre-sterilised soil only > saprotrophs soil < control with Q. ilex soil Plant pathogens -ve correlation in NS NS NS NS NS pre-sterilised and drought soil, +ve correlation in control soils Yeasts +ve correlation in NS NS NS NS soil only < pre-sterilised soil, with Q. ilex no correlation in control soil, -ve correlation in drought soil Table S7. Summary of differences in relative abundance of fungal functional groups contributing > 1.5 % of total reads during the drought experiment. NS, not significant. +ve = positive and -ve = negative. 58 180 pots in total 90 pots soil only 90 pots with plant 30 pots - 30 pots - 30 pots - 30 pots - 30 pots - 30 pots - control soil drought soil sterile soil control soil drought soil sterile soil 10 levels 10 levels 10 levels 10 levels 10 levels 10 levels Divided of drought of drought of drought of drought of drought of drought between (3 reps at (3 reps at (3 reps at (3 reps at (3 reps at (3 reps at six blocks each level) each level) each level) each level) each level) each level) Fig. S1. Outline of the original experimental design, showing number of replicates in each treatment combination. Note that while there were three replicates of each treatment combination (plant × soil history × drought level) there were six experimental blocks. This was due to a second half of the experiment (not used in this article) which looked at the recovery following drought, thus doubling the number of replicates. Fig. S2. Mean soil temperature throughout the experiment measured in five pots (due to availability of soil moisture probes). 59 Fig. S3. Percentage soil moisture for the ten different drought levels measured at the end of each drought treatment. Points are means for each drought level, and shown in dark orange for pots without the plant (‘soil only’) and blue for pots with the plant (‘with Q. ilex’). Standard error bars are shown. n = 18 at each drought measurement for each point, due to a second half of the experiment (not used in this article) which looked at the recovery following drought, thus doubling the number of replicates. (a) Bacteria (b) Fungi a a a a a b Fig. S4. Relationships between (a) bacterial and (b) fungal Shannon diversity and soil history and the presence of Q. ilex, prior to onset of drought (drought duration of zero days). Soil treatments are ‘Control’, ‘Drought’ (pre-droughted) and ‘Sterile’ (pre-sterilised). For bacteria there was a significant interaction (P < 0.001) between soil history and plant. Letters (a and b) signify significant (P < 0.05) differences following a Tukey test. For fungi there was a significant (P < 0.05) soil effect (pre-droughted higher than pre-sterilised). 60 (a) Bacteria (b) Fungi a a ab a ab ab ab a ab ab b b Fig. S5. Relationships between (a) bacterial and (b) fungal Simpson index and soil history and the presence of Q. ilex, prior to onset of drought (drought duration of zero days). Soil treatments are ‘Control’, ‘Drought’ (pre-droughted) and ‘Sterile’ (pre-sterilised). Letters (a and b) signify significant (P < 0.05) differences (within bacteria or fungi) following a Tukey test. (a) (b) Fig. S6. Relationships between the ratio of Gram positive to Gram negative bacteria and (a) drought duration (days) (P < 0.05) and (b) soil history (P < 0.001) and the presence of Q. ilex (P < 0.001). For plot (a) small grey points represent individual samples and larger black points are mean values. For plot (b) soil treatments are ‘Control’, ‘Drought’ (pre-droughted) and ‘Sterile’ (pre-sterilised). 61 (a) (b) a a a a a b Fig. S7. Relationships between bacterial Simpson index and (a) drought duration (days) (P < 0.05) and (b) soil history and the presence of Q. ilex (significant interaction, P < 0.001). Soil treatments are ‘Control’, ‘Drought’ (pre-droughted) and ‘Sterile’ (pre-sterilised). For plot (a) small grey points represent individual samples and larger black points are mean values. (a) (b) a b ab b a a a a a b Simpson index Fig S8. Relationships between fungal Simpson index and (a) drought duration as a categorical variable (Control, 0 days of drought; Low, 2-7 days; Mid, 9-14 days; High, 16-21 days; P < 0.001) and (b) soil history and the presence of Q. ilex (significant interaction, P < 0.01). Letters a and b denote treatments which differ from each other (Tukey HSD test, P < 0.05). Soil treatments are ‘Control’, ‘Drought’ (pre- droughted) and ‘Sterile’ (pre-sterilised). 62