Spatial heterogeneity in plant-soil feedbacks alters competitive interactions between two grassland plant species

The effects of plants on soil vary greatly between plant species and in mixed plant communities this can lead to spatial variation in plant-soil feedback (PSF) effects. Such spatial effects are thought to influence plant species coexistence, but the empirical evidence for this hypothesis is limited. Here, we investigate how spatial heterogeneity in PSFs influences plant growth and competition. The experiment was carried out with high and low nutrient soils to examine how these effects depend on soil fertility. We collected soil from field plots planted for three years with monocultures of Anthoxanthum odoratum and Centaurea jacea and tested the performance of the two species in a greenhouse experiment in heterogeneous soils consisting of patches of “own” and “foreign” soils and in soils where the “own” and “foreign” soils were mixed homogeneously. In the test phase, plants were grown in monocultures and in 1:1 mixtures in live or sterilized soils. Overall, A. odoratum in monocultures produced less aboveground biomass in heterogeneous soils than in homogeneous soils. Centaurea jacea produced less belowground biomass in live heterogeneous soils than in live homogeneous soils, but there was no difference between sterile heterogeneous and homogeneous soils. The belowground biomass per patch varied more in pots with live heterogeneous soils than in pots with live homogeneous soils for both plant species, but there was no difference between pots with sterile heterogeneous and homogeneous soils. In pots with plant mixtures, the difference in aboveground biomass between the two competing species tended to be smaller in heterogeneous than in homogeneous soils. In pots with heterogeneous soils, both plant species grown in mixtures produced more aboveground biomass in “foreign” soil patches than in “own” soil patches. The responses of plants to heterogeneous PSFs were not different between low and high nutrient soils. Our results show that spatially heterogeneous PSFs can influence plant performance and competition via reducing the growth inequality between the two competing species by allowing selective growth in foreign soil patches, independent of initial soil nutrient availability. Such effect may slow down exclusion processes and thus promote the coexistence of competing species at the local scale in mixed plant communities. A plain language summary is available for this article.

Introduction 8 sealed bags. One of the two bags from each plot was sterilized by γ-irradiation (minimum 151 25KGray, Isotron, Ede, the Netherlands). Hence, there were 40 different conditioned soils (2 152 nutrient levels × 2 plant species × 5 replicate plots × 2 sterilization treatments). In the 153 greenhouse experiment, for each of the two nutrient levels and for sterile and non-sterile soil, 154 we created two levels of PSF heterogeneity (spatially homogeneous PSF and spatially 155 heterogeneous PSF) using soils conditioned by A. odoratum and C. jacea from the same field 156 block (Fig. 1). A total of 120 pots (2 nutrient levels × 2 sterilization treatments × 2 PSF 157 heterogeneity treatments (described below) × 3 planting treatments (described below) × 5 158 replicates) of 4.6 L each were used in the greenhouse experiment.

160
Greenhouse experiment 161 In the greenhouse experiment, two levels of PSF heterogeneity (spatially homogeneous PSF 162 and spatially heterogeneous PSF) were created using soil conditioned by A. odoratum and C. 163 jacea from the same field block (Fig. 1). In the heterogeneous soil treatments, each pot was 164 equally divided into 4 patches using a metal grid and each patch was alternately filled with 1.4 165 kg soil conditioned by monocultures of A. odoratum or C. jacea. In the homogeneous soil 166 treatments, each pot was filled with 5.6 kg of a 1:1 (w:w) homogenized mixture of soil 167 conditioned by monocultures of A. odoratum and C. jacea (Fig. 1). In this way, there were 168 pots that differed in spatial variation in plant-soil feedbacks while the abiotic and biotic soil 169 conditions in the homogenous and heterogeneous soils were kept constant. We allocated pots 170 filled with soils originated from the same field block in the same block in the greenhouse 171 experiment so that there were five blocks. Pots of different treatments were randomized 172 within each block. Holes were made in the bottom of each pot to allow vertical movement of 173 water. To prevent soil from passing through holes, a piece of filter paper (15 cm in diameter) 174 Page 8 of 41 Functional Ecology: Confidential Review copy Functional Ecology: Confidential Review copy was placed at the bottom of each pot before filling the pot with soil. Each pot was placed on a 175 tray to prevent possible contamination through leachate. The metal grid was removed after for soil chemical analysis. We measured soil organic matter content, nutrient content (NH 4 ,179 NO 3 and PO 4 ), water content and pH (Table S1). The amount of NH 4 , NO 3 and PO 4 (mg/kg 180 dry soil) were determined by adding 30.0 ml of 0.01 mol/L CaCl 2 solution to soil samples (3.0 181 g), shaking mechanically for at least 2 h at room temperature (20 ℃), filtering the solution and 182 analyzing the nutrients in the soil extracts in a flow analyzer (SKALAR SAN plus system).

183
Soil pH-H 2 O was determined by adding 25.0 ml demi-water to soil samples (volume 5.0 ml), 184 shaking for 5 min and measuring 2 h later. Soil organic matter was determined by measuring 185 the difference between weights of the oven-dried (105 ℃) soil samples (5.0-10.0 g) before 186 and after being heated in a furnace at 550 ℃. The weight of each sample was determined after 187 cooling it down in the air to handwarm temperature and further cooling it for at least 45 min 188 in a desiccator. Soil moisture content was determined by measuring the difference between 189 the weights of each soil samples before and after oven-drying (105 ℃).

190
In a heated greenhouse (20.0 ℃ average temperature, 70.2 % average relative humidity), 191 seeds of A. odoratum and C. jacea (purchased from a wild seed supplier, Cruydthoeck,192 Nijeberkoop, the Netherlands) were sown on plastic trays filled with steamed potting soil that Lawn Products, Inc., Marysville). The potting soil was watered daily so that the potting soil 195 remained moist. One week after germination, the trays with seedlings were moved to an 196 unheated greenhouse (12.8 ℃ average temperature, 70.3 % average relative humidity) until 197 they were transplanted into the pots.

198
Page 9 of 41 Functional Ecology: Confidential Review copy Functional Ecology: Confidential Review copy Similar sized seedlings of A. odoratum and C. jacea were used in the experiment. There were 199 three planting treatments, i.e. the two species were planted in monocultures and in 1:1 200 mixtures ( Fig. 1). In monocultures, we planted 16 seedlings (a similar planting density as 201 applied in Wubs & Bezemer, 2016) of A. odoratum or C. jacea in each pot. In mixtures, we 202 planted eight seedlings of A. odoratum and C. jacea in alternating positions (Fig. 1). In this 203 way, each seedling was surrounded by conspecific and heterospecific competitors. Dead 204 seedlings were replaced during the first week of the experiment. We removed the dead 205 seedlings, including the root system, and then planted a new seedling at the previous planting 206 position. All other species emerging from the seed bank of the soil were removed manually 207 during the experiment.

208
The experiment was maintained for 90 days (from 11 April to 11 July 2016) in the same 209 unheated greenhouse. During the experiment, the mean temperature and the relative humidity 210 in the greenhouse were 17.4 ℃ and 67.5 %, respectively. All pots were watered three times 211 per week (300-800 ml per pot, each time depending on the weather conditions).

212
In this experiment, we analysed the effects of spatial plant-soil feedback heterogeneity by 213 comparing spatially heterogeneous soils with homogeneously mixed soils that have the same 214 origin. Hence, each pot consisted of the same initial nutritional and microbial composition.

215
For completeness, in the experimental design we also included the two pure soil treatments 216 (pure Ao soils and pure Cj soils; Fig. 1). In these two pure soil treatments, each pot was filled 217 with 5.6 kg of soil conditioned by monocultures of A. odoratum (pure Ao soil treatment) or C. 218 jacea (pure Cj soil treatment) growing in either high or low nutrient soil and originating from 219 the same field block. The data of root and shoot biomass in these pure soils are presented in 220 the supplementary information ( were harvested separately. In the 1:1 mixtures, the two different species were also harvested 225 separately. After clipping, we took one soil core (4.0 cm diameter, straight down to the 226 bottom of pot) in each of the four soil patches in each pot to measure the root mass (Fig. 1).

227
Soil cores were only taken from pots planted with monocultures since it was not possible to    Then, we analyzed the patch-level aboveground biomass and belowground biomass separately 258 using a mixed-effect three-way ANOVA to test whether the two species grown in 259 monocultures produced more biomass in "foreign" soil patches than in "own" soil patches 260 within the heterogeneous soil. In this model, nutrient availability, sterilization, soil type ("own" 261 vs. "foreign" soil) and their interactions were included as fixed factors, soil type nested in pot, 262 and pot nested in block (block/pot/soil type) was included as a random effect to account for 263 the non-independent of the growth in different patches within one pot.

264
For mixed plant communities, at the pot level, we first combined the growth of the two 265 species in 1:1 mixtures in each pot by calculating the growth difference (D) to evaluate the 266 effects of spatial plant-soil feedback heterogeneity on the competition between the two 267 species. The D-value was calculated as the log-ratio of aboveground biomass of A. odoratum 268 and C. jacea in mixtures. The D-value will be equal to zero if the two species perform equally 269 well in mixtures; it will be positive if the biomass of A. odoratum is higher than C. jacea, and 270 negative if C. jacea biomass is higher. We used three-way ANOVA to test the effects of

282
At the patch level, we tested whether the two species in the 1:1 mixtures produced more 283 biomass in "foreign" soil patches than in "own" soil patches within the heterogeneous soils. 284 We analysed the patch-level aboveground biomass (total aboveground biomass of a species in 285 one patch divided by the number of seedlings in the patch) separately for each of the two 286 species grown in the 1:1 mixture, using a mixed-effect three-way ANOVA. Nutrient 287 availability, sterilization, soil heterogeneity and their interactions were included as fixed 288 factors, and soil type nested in pot, and pot nested in block (block/pot/soil type) as a random

Effects of plant-soil feedback heterogeneity on the growth in monocultures 297
In monocultures, A. odoratum overall produced less aboveground biomass in heterogeneous 298 soils than in homogeneous soils (Table S3A; Fig. 2A), but there was no significant difference 299 in the aboveground biomass of C. jacea between the two soils (Table S3A; Fig. 2C). These 300 results suggest that heterogeneity in PSFs did influence the aboveground biomass of A. 301 odoratum but not of C. jacea. Both species produced much more aboveground biomass in 302 sterile soil than in live soil (Table S3A; Fig

304
PSF heterogeneity also influenced belowground biomass but the effect varied between the 305 two species and soil sterilization. A. odoratum produced similar amounts of belowground 306 biomass in heterogeneous and homogeneous soils (Table S3B; Fig. 2B). C. jacea produced 307 less belowground biomass in live heterogeneous than in live homogeneous soils, but in 308 sterilized soil there was no difference between these heterogeneity treatments (Table S3B: 309 significant sterilization × heterogeneity effect; Fig. 2D). These results suggest that 310 heterogeneity in PSFs influenced the belowground biomass of C. jacea but not of A. 311 odoratum. Belowground biomass per soil core of both species was significantly greater in 312 sterile soil than in live soil (Table S3B; Fig. 2B, D).

313
Soil heterogeneity and the interaction with nutrient and/or sterilization did not affect the CV 314 of aboveground biomass of either A. odoratum or C. jacea (Table S4A; Fig. 3A, C). CVs of 315 belowground biomass of both plant species were significantly greater in live heterogeneous 316 soil than in live homogeneous soil. In sterilized soil there was no difference between the two 317 heterogeneity treatments (Table S4B:  Functional Ecology: Confidential Review copy Functional Ecology: Confidential Review copy heterogeneity increased spatial variation in root growth in live soil but not when soil biota 320 were excluded.

321
In monocultures, in pots with spatially heterogeneous soil, A. odoratum produced more 322 aboveground biomass in live "foreign" soil patches than in live "own" soil patches when soil 323 nutrient is low, but no difference was found between these two patches in high nutrient soil or 324 in sterile soils (Table S5A: significant nutrient × sterilization × soil interaction effect; Fig.   325 S3A). C. jacea produced more aboveground biomass in live "foreign" soil patches than in live 326 "own" soil patches but there was no difference between the two soil patches in sterile soils 327 (Table S5A: significant sterilization × soil interaction effect; Fig. S3C). The same pattern was 328 found for the belowground biomass of A. odoratum, while C. jacea overall produced less 329 belowground biomass in "foreign" soil patches than in "own" soil patches (Table S5B;   In mixtures, the growth difference between the two species tended to be smaller in 335 heterogeneous soils than in homogeneous soils (Table S6: marginally significant 336 heterogeneity effect; Fig. 4), indicating that the growth inequality between the two competing 337 species was reduced in heterogeneous soils. The growth difference index (D) was generally 338 negative in live soil but positive in sterile soil, i.e. C. jacea was superior to A. odoratum in 339 live soil while the reverse was true in sterile soil (Table S6; Fig. 4). The aboveground biomass 340 of both species grown in mixtures is presented in the supporting information ( In mixtures, in pots with spatially heterogeneous soil, A. odoratum produced more 343 aboveground biomass in "foreign" soil patches than in "own" soil patches (Table S5C; Fig.   344 5A). A similar trend was observed for C. jacea but this was not significant (Table S5C;   our study, we only found weak evidence for this. In heterogeneous soils, both plant species 395 encountered patches with "own" and "foreign" soils, potentially providing both plant species 396 with enemy free space, i.e. the avoidance of contact with antagonists in "own" soil patches.

397
Indeed in mixtures, we generally found a negative conspecific PSF (less growth in "own" than 398 in "foreign" soil patches) even though this was only significant for one of the two species.

399
This result indicates that spatially heterogeneous PSFs can reduce the biomass inequality 400 between competing species but also shows that the effects are plant species specific.

401
As expected, sterilizing the soil increased plant growth. Our results show that soil biota in our 402 system have a negative effect on plant growth, i.e. there are more pathogenic or harmful 403 microbes than beneficial ones present in conditioned soil. However, it is important to note 404 sterilization of soils also increased the soil nutrient availability (Table S1) , 1997). In contrast to our hypothesis, the effects of PSF heterogeneity did not differ 431 between the two soil fertility levels as indicated by the absence of significant nutrient × 432 heterogeneity effects. At the end of the conditioning period in the field, the amount of organic 433 matter was higher in high nutrient than in low nutrient soils, but there were no differences in 434 other soil chemical properties between the two soil nutrient treatments (Table S1). This may 435 explain why we did not observe stronger conditioning effects on PSF heterogeneity effects in     monocultures among the four patches within homogeneous and heterogeneous soils. "High" 697 and "Low" refer to high nutrient soil and low nutrient soil used in the conditioning phase.

698
"Live soil" and "Sterile soil" indicate field-collected soil and sterilized field-collected soil, 699 respectively. Mean values (± 1 SE) are presented. See Table S4 for statistic results. biomass of A. odoratum is higher than C. jacea and negative values indicate the reverse is true.

705
"High" and "Low" refer to high nutrient soil and low nutrient soil used in the conditioning 706 phase. "Live soil" and "Sterile soil" indicate field-collected soil and sterilized field-collected 707 soil, respectively. Mean values (± 1 SE) are presented. See Table S6 for statistic results. Stars mixtures in "own" and "foreign" soil patches for pots with heterogeneous soils. "High" and 714 "Low" refer to high nutrient soil and low nutrient soil used in the conditioning phase. "Live 715 soil" and "Sterile soil" indicate field-collected soil and sterilized field-collected soil, 716 respectively. Mean values (± 1 SE) are presented. See Table S5C for statistic results.       are for monocultures with heterogeneous soils. "Own" and "foreign" soil patches refer to 59 conspecific and heterospecific soil patches respectively. "High" and "Low" refer to high 60 nutrient soil and low nutrient soil used in the conditioning phase. "Live soil" and "Sterile soil" 61 indicate field-collected soil and sterilized field-collected soil, respectively. Mean values (± 1 62 SE) are presented. See Table S5A-B for statistic results.  in homogeneous and heterogeneous soils at the pot level. "High" and "Low" refer to high 68 nutrient soil and low nutrient soil used in the conditioning phase. "Live soil" and "Sterile soil" 69 indicate field-collected soil and sterilized field-collected soil, respectively. Mean values (± 1 70 SE) are presented. See Table S3C for statistic results.