Elevated tropospheric CO2 and O3 concentrations impair organic pollutant removal from grassland soil

The concentrations of tropospheric CO2 and O3 have been rising due to human activities. These rising concentrations may have strong impacts on soil functions as changes in plant physiology may lead to altered plant-soil interactions. Here, the effects of eCO2 and eO3 on the removal of polycyclic aromatic hydrocarbon (PAH) pollutants in grassland soil were studied. Both elevated CO2 and O3 concentrations decreased PAH removal with lowest removal rates at elevated CO2 and elevated O3 concentrations. This effect was linked to a shift in soil microbial community structure by structural equation modeling. Elevated CO2 and O3 concentrations reduced the abundance of gram-positive bacteria, which were tightly linked to soil enzyme production and PAH degradation. Although plant diversity did not buffer CO2 and O3 effects, certain soil microbial communities and functions were affected by plant communities, indicating the potential for longer-term phytoremediation approaches. Results of this study show that elevated CO2 and O3 concentrations may compromise the ability of soils to degrade organic pollutants. On the other hand, the present study also indicates that the targeted assembly of plant communities may be a promising tool to shape soil microbial communities for the degradation of organic pollutants in a changing world.


Results
PAH residuals in soil. Both eCO 2 and eO 3 increased total PAH residuals significantly (Table 1, Figs 1

and 2),
i.e. decelerated PAH degradation, and also altered the composition of remaining PAHs (increased PC1 of PAHs; Fig. 2). CO 2 × O 3 had a significant interactive effect on total PAH residuals as remaining PAHs were lowest at aCO 2 and aO 3 , but substantially increased by eCO 2 and eO 3 and highest at both eCO 2 and eO 3 (+43% in comparison to ambient conditions). Plant functional group richness had no significant effects on PAH residuals ( Table 1). The PLFAs i17:0, cy 17:0, and i16:0 were most strongly associated with PAH removal, and Benzo(k)fluoranthene and Indene(1,2,3-c,d)pyrene were the most recalcitrant PAHs (Fig. 3). Gram-negative bacteria were positively Error df F-value P-value F-value P-value F-value P-value F-value P-value F-value P-value F-value P-value F-value P-value associated with total PAH residuals and PC1 of PAH residuals, while Gram-positive bacteria strongly reduced total PAHs and PC1 of PAH residuals (Fig. 2). Although the fungal PLFA 18:2ω6, 9 was very abundant in the experimental soil (Fig. S3), it played a minor role in PAH degradation (Fig. 3). Furthermore, Gram-positive bacteria were the most important group of soil microbes in degrading PAHs (Figs 2 and 3).
Soil enzymes and microorganisms. Both eCO 2 and eO 3 significantly reduced polyphenol oxidase activity, but plant functional group richness increased polyphenol oxidase activity (Table 1, Fig. S4b). However, enzyme Figure 1. Total amount of polycyclic aromatic hydrocarbons (PAHs) after the experiment as affected by elevated CO 2 , elevated O 3 , and plant diversity (0, 1, 2, 3 plant functional groups). Means ± SE (n = 4). amb, eCO 2 , eO 3 , and eCO 2 + eO 3 means that microcosms were incubated in chambers with ambient air, with elevated CO 2 , with elevated O 3 , and with elevated CO 2 and O 3 , respectively. Bars with different letters vary significantly (Tukey's HSD test, a <0.05).  3 , and plant functional group richness on soil microorganisms and polycyclic aromatic hydrocarbon (PAHs) residuals in the soil. Red arrows: negative relationships, blue arrows: positive relationships, asterisks on numbers indicate significant relationships (see Table S1 for details).
Furthermore, phenol oxidase activity was marginally affected by eO 3 in the presence of 0 and 1 plant functional group, but increased phenol oxidase activity in the presence of 2 and 3 plant functional groups (significant O 3 × plant functional group richness interaction). Phenol oxidase activity decreased with increasing plant functional group richness and was consistently higher at eCO 2 and aO 3 , but did not vary with plant functional group richness and CO 2 at eO 3 (significant CO 2 × O 3 × plant functional group richness interaction). Moreover, the increase of polyphenol oxidase activity at eCO 2 was most pronounced in the presence of two plant functional groups (significant CO 2 × plant functional group richness interaction).
Elevated CO 2 significantly reduced the biomass of Gram-positive bacteria ( Table 1, Fig. S1a), and the composition of microbial communities changed as indicated by increased PC1 of microbial PLFAs (Table S1, Fig. 2). Furthermore, eO 3 significantly reduced Gram-positive bacteria and Gram-negative bacteria (Tables 1, S1, Figs 2, S1a,b) and changed the composition of the soil microbial community (increased PC1 of microbial PLFAs; Table S1, Fig. 2). The biomass of Gram-positive bacteria, Gram-negative bacteria, and fungi increased significantly with increasing plant functional group richness (Table 1, Fig. S1a,b,c), which was also reflected by a significant change of soil microbial community composition (significantly reduced PC1 of microbial PLFAs; Table S1, Fig. 2). CO 2 × O 3 and CO 2 × plant functional group richness had significant interactive effects on Gram-positive bacteria ( Table 1). The biomass of Gram-positive bacteria was highest at aCO 2 and aO 3 , lowest at aCO 2 and eO 3 , and intermediate in the other treatments (Fig. S1a). Further, the biomass of Gram-positive bacteria increased with increasing plant functional group richness and was higher at aCO 2 , except in the treatment with two plant functional groups, where aCO 2 and eCO 2 had similar values.

Discussion
The present study shows that elevated CO 2 and O 3 concentrations may erode essential ecosystem services like the degradation of pollutants in soil by inducing significant shifts in soil microbial community structure and enzyme activity. These detrimental effects were consistent across plant communities differing in functional diversity. However, pronounced alterations of microbial community structure along the functional plant diversity gradient suggests that targeted and trait-based phytoremediation may help to counteract detrimental global change effects in long-term approaches. In contrast to our hypothesis (1) stating that plant functional group richness to increase plant productivity 37 , plant biomass declined with increasing plant functional group diversity in the present study. These results highlight the context-dependency of biodiversity-ecosystem function relationships 43 and the need to study biodiversity effects under stressful conditions 44 . Nevertheless and in line with our hypothesis 17,38 , soil microbial biomass and activity increased with plant functional group diversity, stressing the significance of plant diversity for soil functions. Diverse plant communities are expected to produce and release a higher quantity and diversity of organic compounds into their rhizosphere, which may sustain higher soil microbial biomass and activity 45 . Using long-term data from a grassland biodiversity experiment, Lange et al. found higher plant biodiversity to increase rhizosphere carbon inputs into the soil microbial community resulting in increased microbial diversity and activity 18 . These findings are consistent with the results of increased biomass of Gram-positive, Gram-negative bacteria and fungi in the present study. A recent meta-analysis reported that plant diversity effects on soil microbial biomass C were strong in long-term experiments and across various environmental contexts 46 . The present study extends those findings by showing that bacterial and fungal biomass increased with plant diversity, which also altered the activities of different soil enzymes. Similar with this study, Steinauer et al. found that soil microbial biomass and some enzyme activities increased with increasing plant diversity 47 .
In contrast to our hypotheses, no effect of plant functional group richness on PAH removal was observed. This is in line with a previous study that species richness had no significant effect on 14 C-phenanthrene mineralization 48 . However, structural equation modeling (SEM) reveals a range of processes coupling plant functional group richness and PAH degradation. The SEM showed that plant diversity altered soil microbial community composition and favored both Gram-positive (accelerating PAH degradation) and Gram-negative bacteria (decelerating PAH degradation; Fig. 2). Plant community may therefore be an important driver of PAH degradation, even if lumping community composition into functional group richness doesn't provide the adequate explanatory power. The present study suggests that it may be possible to assemble plant communities showing a high phytoremediation by steering soil microbial communities.
Elevated CO 2 tended to increase plant productivity, although the results were only marginally significant. In addition, eCO 2 had a negative impact on microbial processes linked to PAH degradation. Although eCO 2 increased total soil microbial biomass and activity (Fig. 2, PC1 microbial PLFAs), it led to a decrease in Gram-positive bacteria, a microbial group linked to PAH degradation 49 and the most important microbial group involved in the removal of PAHs from soil in the present study (Fig. 2). Furthermore, eCO 2 altered the soil microbial community composition, which is also in line with previous studies 38,50,51 and calls for more detailed investigations of shifts in soil microbial communities with sequencing techniques.
We propose that this effect of elevated tropospheric CO 2 may be due to the higher plant carbon input in soil resulting from enhanced photosynthesis 52 . This may lead to higher soil microbial activity 29,53 , as the pool of labile soil C may be increased by elevated root exudation 33,54,55 . In the present study, eCO 2 had non-significant effects on the biomass of fungi and Gram-negative bacteria, but decreased the biomass of Gram-positive bacteria (Fig. S1). Consistent with the present study, both Larson et al. 31 and Grueter et al. 56 found that eCO 2 had no significant influence on microbial biomass and activity, while Manninen et al. 57 found a negative effect of eCO 2 on soil microbial biomass. These variable results indicate that eCO 2 effects on soil microbial communities may depend on the environmental context, such as soil conditions and/or plant community composition 38 .
Elevated O 3 decreased plant biomass, soil microbial biomass and activity, and PAH removal. Ozone is a toxic compound that can induce oxidative stress in plants, and high tropospheric O 3 concentrations have been reported to decrease inputs and to change the composition of assimilates into the rhizosphere 34 , which in turns affects soil microbial communities. Results of the present study indicate that ozone-mediated changes in soil communities may have dramatic effects on soil self-cleaning potential. Consistent with past studies 58-60 , a strong decrease in the biomass of Gram-positive and Gram-negative bacteria and shifts in microbial community composition in response to eO 3 was observed (Figs 2, S1).
The effects of eCO 2 , eO 3 , and plant diversity on PAH removal were mediated to some extent by alterations of soil enzymatic activity. Both elevated CO 2 and O 3 led to a decrease in polyphenol oxidase activity, while plant functional group richness increased polyphenol oxidase but decreased phenol oxidase. In line with the present study, eCO 2 reduced the activity 61 and abundance 62 of polyphenol oxidase, suppressed phenol oxidase 63 , while enzymes including phenol oxidase were strongly affected by plant species richness 64 . These results indicate that simultaneous alteration of plant community composition and environmental conditions may have contrasting effects on enzyme activity involved in PAH removal. Notably, many enzymes are involved in the metabolism process of PAHs 65,66 , some of which were not measured here. Although the measured enzymes responded significantly to the treatments, this did not explain variation in PAH removal, which is why they were not considered in the structural equation model (Fig. 2).
Elevated CO 2 and O 3 concentrations and variations in plant diversity had significant interactive effects on plant biomass, soil microbial functions, and the degradation of PAHs. Plant diversity altered the effect of eCO 2 on soil microbial biomass and activity, but the clear positive interaction effects as expected in hypothesis (4) were not detected. This highlights the importance of plant diversity and community composition in mediating soil microbial functions in a future world, but also calls for a better mechanistic understanding of interactive effects of plant diversity and global change drivers.
However, plant diversity did not alter eCO 2 and eO 3 effects on PAH removal in the present study. This is in line with a recent meta-analysis by Thakur et al. 46 showing no interactive effects of plant diversity and global change factors in affecting soil microbial biomass in the short term. Potentially, plant diversity-induced differences in soil microbial community composition and subsequent effects on essential services like PAH degradation need a longer time than captured by the present experiment 46 . Moreover, we propose that lumping plant community composition into functional group richness may not provide the adequate explanatory level. Instead, we propose that future studies may use more targeted plant trait-based approaches 67  Importantly and in contrast to our hypothesis (4), eCO 2 amplified the inhibitory effect of eO 3 on PAH removal. This effect was partly mediated by an enhancement of eO 3 effects on most soil microbial groups at elevated CO 2 . It particularly amplified the negative effect of eO 3 on Gram-positive bacteria, the most important microbial group driving the removal of PAHs from soil in this study. This result exemplifies how different global change drivers can have unexpected synergistic effects on soil functions and compromise important ecosystem services.

Conclusion
We highlight that global environmental change factors, such as human-induced alterations in tropospheric gas composition, may undermine the ability of ecosystems to degrade pollutants. Soil self-cleaning showed a high robustness to alterations in plant diversity and community composition, yet elevated CO 2 and O 3 concentrations may compromise efforts such as phytoremediation to restore polluted soils. On the other hand, the present study also indicates that the targeted assembly of plant communities applying a more comprehensive knowledge regarding plant effects on soil biota may be a promising tool to shape soil microbial communities for the degradation of organic pollutants.

Materials and Methods
Open top chambers. The open top chamber (OTC) system is located at Xianlin campus, Nanjing University, Nanjing, China (118°57′36.15″E, 31°7′23.99″N). Briefly, this system consists of four chambers with full control of atmospheric CO 2 and O 3 concentrations: one chamber with ambient CO 2 (aCO 2 ) and ambient O 3 (aO 3 ) levels, one with eCO 2 and aO 3 levels, one with aCO 2 and eO 3 levels, and one with both eCO 2 and eO 3 levels. The glass chambers are octagonal with 2 m in diameter and 2.8 m in height. CO 2 was released from a tank (Q/JB-THB002, Beijing Tianhai Industry Co., Ltd.), and O 3 was produced by an O 3 generator (NPF10/W, Shandong Lvbang Ozone Co., Ltd.) from pure O 2 . CO 2 and/or O 3 were mixed with air from temperature-controlled rooms and conveyed by fans (SFG-2, Shanghai Jiabao Co., Ltd.) to the bottom of the chambers. Gases were released into the antra via tiny holes in the stainless steel plate between the bottom and the antrum, and then released into the air of the open top of chambers. The quantity of the CO 2 and O 3 release was controlled by a flowmeter (LZB-3WB, Changzhou Shuangbo Co., Ltd.), the concentration of CO 2 was detected with a CO 2 monitor (Li-7000, Li-Cor, USA), and the concentration of O 3 was detected with an O 3 monitor (Model 205, 2B Co., USA). The O 3 fumigation was conducted between 9:00 a.m. and 5:00 p.m. until harvest, except during rain events, and the CO 2 fumigation was all day long until harvest. The target CO 2 concentration for the eCO 2 treatment was 200 ppm higher than aCO 2 , and the target O 3 concentration for the eO 3 treatment was 50-60 ppb higher than aO 3 in order to simulate the forecasted tropospheric CO 2 and O 3 levels in 2050 2 .
Plant cultivation. Three species of grasses (Lolium perenne, Dactylis glomerata, Phleum pratense), herbs (Plantago lanceolata, Taraxacum officinale, Centaurea jacea), and legumes (Trifolium pratense, Trifolium repens, Medicago sativa) were germinated in trays filled with quartz sand in the lab. Ten days after germination, seedlings were transplanted into the microcosms (8 cm in diameter and 12 cm in height) with 250 g of PAHs contaminated soil collected from a chemical plant in Nanjing (118°44′51.87″E, 31°58′4.71″N). Plant communities consisting of nine individuals and differing in functional group richness (8 different communities) were set up: bare ground (no plants); functional group 'monocultures' of either three grass species, three herb species, or three legume species; mixtures of two functional groups (grasses plus herbs, grasses plus legumes, or herbs plus legumes); and the mixture containing all three plant functional groups (grasses plus herbs plus legumes), thereby yielding functional group richness levels of 0, 1, 2, and 3 and functionally dissimilar plant communities. Each plant community was replicated four times per CO 2 × O 3 treatment (32 microcosms per OTC, 128 microcosms in total). Plant communities were cultivated in the lab for one week, and dead seedlings were replaced before microcosms were transferred to OTCs.
The microcosms were randomly placed in the OTCs, and each microcosm was watered with 10-20 ml of distilled water per day. After 10 weeks of cultivation in OTCs, plants and soils were sampled, survival of plants and plant community biomass was measured, and soils for PAHs determination were stored at −20 °C, whereas soils for the measurement of microbial parameters were stored at 4 °C.

Determination of soil enzymatic activity. A very important step of PAH metabolism by bacteria and
fungi is the breaking of PAH rings by phenol oxidase or polyphenol oxidase 65,66 . Therefore, these enzymes were used as proxy for general microbial processes linked to PAH degradation. For enzyme measurements, 0.5 g fresh soil was mixed with 20 ml milli-Q-water in 50 ml falcon tubes, shaken at 250 rpm for 30 min, centrifuged at 3000 rpm for 10 min, supernatants mixed with substrates and buffer in 96-well plates (Corning 96 Flat Bottom Transparent Polystyrol), then determined on a plate reader (Infinite M200, Tecan, Germany). Phenol oxidase activity was measured according to a modified protocol 68  and split up into phospholipids by eluting with chloroform, acetone, and methanol from a silica-filled solid phase extraction column. Subsequently, the phospholipids were hydrolyzed and methylated by a methanolic KOH solution, and the PLFA-methyl esters were identified and quantified by GC-ECD (PerkinElmer, Clarus 500, USA). PLFA 19:0 was used as internal standard. Separated phospholipid fatty acid methyl-esters were identified by chromatographic retention time and mass spectral comparison with a mixture of standard qualitative bacterial acid methyl-ester that ranged from C11 to C24 (Supelco). For each sample, the abundance of individual phospholipid fatty acid methyl-esters was expressed in nmol per g dry soil. The nomenclature for PLFAs followed that of Frostegård et al. 72 . The sum of PLFAs i14:0 i15:0, a15:0, i16:0, i17:0, and i18:0 represented the biomass of Gram-positive bacteria, that of PLFAs cy17:0 and cy19:0 represented the biomass of Gram-negative bacteria, and the amount of the fungal-specific fatty acid 18:2ω6,9 was used as an indicator of fungal biomass 73,74 . Determination of PAHs in soils. Soil samples stored at −20 °C were freeze-dried (Labconco 12 L, Labconco Co., USA) for 96 h, ground by mortars, and passed through a 2 mm sieve. Samples (5 g) were extracted with 20 mL methanol: methylene dichloride (1:2, v-v), concentrated in a rotary evaporator, and dried under a fine stream of nitrogen. The residues were dissolved in 0.5 ml acetonitrile. Samples were analyzed by high-performance liquid chromatography on a Spuelcosil TM LC-PAH column (250 × 4.6 mm, 5 μm) (Supelco, Bellefonete, PA, USA) with UV detector at 254 nm (HPLC-UV, Hitachi L2000). The temperature of the column was kept constant at 30 °C to obtain reproducible retention times. The mobile phase consisted of water and acetonitrile in gradient mode at flow rate of 1 ml/min. The gradient solvent system started with 60% acetonitrile in water (v/v) during 10 min, then increasing linearly to 100% acetonitrile within 10 min, the 100% acetonitrile was maintained for 20 min, and finally returned to the initial conditions in 2 min.

Statistical analyses.
Analyses of variance (ANOVAs) were performed to test effects of CO 2 (ambient and elevated), O 3 (ambient and elevated), plant functional group richness (1, 2, 3 functional groups present), and all interactions on total plant biomass, plant shoot biomass, plant root biomass, plant survival, biomass of Gram-positive bacteria, biomass of Gram-negative bacteria, biomass of fungi, phenol oxidase activity, polyphenol oxidase activity, and total PAH residuals (for the latter, treatments with 0, 1, 2, and 3 functional groups were considered). If significant treatment effects were detected, additional Tukey's HSD tests were performed to test for differences among means. ANOVAs were performed using Statistica 7.1 (Statsoft). Furthermore, structural equation modeling (SEM) was used to shed light on the mechanisms of PAH degradation by accounting for multiple potentially correlated effect pathways to disentangle the direct and indirect effects 75 of experimental treatments and soil microbial community properties. The initial model was based on previous knowledge with experimental treatments as exogenous variables and the endogenous variables "Gram-negative bacteria", "Gram-positive bacteria", "PC1 microbial PLFAs" (representing PLFA composition), "PC1 PAHs" (representing PAH composition), and "total PAHs". The adequacy of the models was determined via chi²-tests, AIC, and RMSEA 76 . Model modification indices and stepwise removal of non-significant relationships were used to improve the models; however, only scientifically sound relationships were considered 75 . Structural equation modeling was performed using Amos 5 (Amos Development Corporation, Crawfordville, FL, USA). Data availability. All data generated and/or analyzed during the current study are available from the corresponding author on reasonable request.