Plant community diversity alters the response of ecosystem multifunctionality to multiple global change factors

Biodiversity is considered important to the mitigation of global change impacts on ecosystem multifunctionality in terrestrial ecosystems. However, potential mechanisms through which biodiversity maintains ecosystem multifunctionality under global change remain unclear. We grew 132 plant communities with two levels of plant diversity, crossed with treatments based on 10 global change factors (nitrogen deposition, soil salinity, drought, plant invasion, simulated grazing, oil pollution, plastics pollution, antibiotics pollution, heavy metal pollution, and pesticide pollution). All global change factors negatively impacted ecosystem multifunctionality, but negative impacts were stronger in high compared with low diversity plant communities. We explored potential mechanisms for this unexpected result, finding that the inhibition of selection effects (i.e., selection for plant species associated with high ecosystem functioning) contributed to sensitivity of ecosystem multifunctionality to global change. Specifically, global change factors decreased the abundance of novel functional plants (i.e., legumes) in high but not low diversity plant communities. The negative impacts of global change on ecosystem multifunctionality were also mediated by increased relative abundance of fungal plant pathogens (identified from metabarcoding of soil samples) and their negative relationship with the abundance of novel functional plants. Taken together, our experiment highlights the importance of protecting high diversity plant communities and legumes, and managing fungal pathogens, to the maintenance of ecosystem multifunctionality in the face of complex global change.

tics pollution, antibiotics pollution, heavy metal pollution, and pesticide pollution).All global change factors negatively impacted ecosystem multifunctionality, but negative impacts were stronger in high compared with low diversity plant communities.We explored potential mechanisms for this unexpected result, finding that the inhibition of selection effects (i.e., selection for plant species associated with high ecosystem functioning) contributed to sensitivity of ecosystem multifunctionality to global change.
Specifically, global change factors decreased the abundance of novel functional plants (i.e., legumes) in high but not low diversity plant communities.The negative impacts of global change on ecosystem multifunctionality were also mediated by increased relative abundance of fungal plant pathogens (identified from metabarcoding of soil samples) and their negative relationship with the abundance of novel functional plants.
Taken together, our experiment highlights the importance of protecting high diversity plant communities and legumes, and managing fungal pathogens, to the maintenance of ecosystem multifunctionality in the face of complex global change.

K E Y W O R D S
biodiversity, complementarity effects, ecosystem functioning, fungal pathogens, legumes, novel functional plants, selection effects complex (Custer & Dini-Andreote, 2022).Although recent studies have shown that many ecosystem functions are inhibited by global change factors (Giling et al., 2019;Zhou et al., 2023), some ecosystem functions are promoted by other global change factors.
For example, increased temperature, nitrogen deposition, and CO 2 concentration have positive impacts on aboveground primary productivity (Su et al., 2022;Zhang et al., 2022).One approach to deal with these complexities and variable impacts is through the consideration of ecosystem multifunctionality, which encompasses the overall response of multiple ecosystem functions (Boyd et al., 2018).However, few studies have used controlled experiments to compare how ecosystem multifunctionality changes in response to multiple individual or interacting global change factors (Rillig et al., 2019), an important step to increase understanding of the relative severity of impacts of multiple global change factors on overall ecosystem functioning (multifunctionality), as well as underlying mechanisms.
Selection effects occur because high plant diversity increases the likelihood that a plant species associated with high ecosystem functioning will occur in the community (Loreau, 1998;Tilman et al., 1997;Tscharntke et al., 2012).However, whether changes in complementarity and/or selection effects underly how diversity mediates the impact of global change factors on ecosystem multifunctionality has not been tested.Furthermore, complementarity and selection effects are also influenced by global change factors (Fanin et al., 2018;Xi et al., 2022), but the underlying mechanisms of these effects are unclear.On the one hand, global change might strengthen complementarity effects and ecosystem multifunctionality by providing novel niches, increased niche width, and decreased niche overlap (Gao et al., 2021;Sardans et al., 2021;Searle & Chen, 2020; Figure 1a).For example, nitrogen deposition provides additional resources for plants, enlarging the resource niche to alleviate interspecific competition (Granjel et al., 2023), decrease niche overlap (Zheng et al., 2022), and maintain ecosystem functioning (Song et al., 2021).On the other hand, global change factors could weaken complementarity effects and ecosystem multifunctionality by reducing available niche space and increasing plant competition for limited resources (Lefcheck et al., 2015;Pires et al., 2018;Figure 1b).Regarding selection effects, high plant diversity increases the probability of a species being present that can support high ecosystem multifunctionality (van der Plas, 2019).Moreover, any species that disproportionately contributes to differences in ecosystem multifunctionality between communities subjected or not to global change factors can be considered a "novel functional species," which might play an important role in selection effects under global change (Jia et al., 2020).If these novel functional species are resistant to global change factors, then this could suggest that differences in ecosystem multifunctionality between communities with low and high biodiversity are based on selection effects (Figure 1a).At the same time, because high diversity communities have a higher probability of containing these novel functional species than low diversity communities, this might mean that the resistance of ecosystem multifunctionality to global change factors is also higher.Conversely, if novel functional species are sensitive to global change factors, the result would be the opposite (Isbell et al., 2011), with ecosystem multifunctionality more sensitive to global change factors in high diversity communities.Therefore, it remains unclear whether high community diversity confers resistance or sensitivity of ecosystem multifunctionality to global change factors, whether this occurs via changes in complementarity or selection effects, and the potential mechanisms underlying these changes (Hong et al., 2022).
Moreover, fungal plant pathogens can mediate the relationship between plant diversity and ecosystem functioning via their strong impacts on plant fitness and resulting changes in relative abundances of plant species in plant communities (Yang, Liang, et al., 2021).If global change alters the relative abundance of fungal plant pathogens, and this in turn impacts plant groups that provide high ecosystem functioning and/or resistance to global change factors, ecosystem multifunctionality may also be impacted (Jia et al., 2020).Therefore, changes in the relative abundance of fungal plant pathogens might be a previously overlooked yet important mediator of the relationship between biodiversity and ecosystem multifunctionality under global change.
We conducted a greenhouse experiment where low (four spe- niche width index (Levins, 1968) and the Schoener niche overlap index (Schoener, 1968)

| Location and study species
The experiment was carried out in a greenhouse at Baimiao Experimental Station, Shandong University, China (36°23′10″ N, 120°36′44″ E), with a daily mean temperature of 30°C and air humidity maintained at ~85% by ventilating fan.We selected two representative species from each of four plant families that commonly co-occur in the Yellow River Delta (Yi et al., 2021): plant species were collected from the Yellow River Delta Field Experiment Station (37°51′9″ N, 118°48′53″ E) during September-October 2020.Seeds were stored at 4°C in a seed storage cabinet (CZ-300FC, TOP Instruments Ltd., China) until the experiment began.

| Experimental design
Plant communities were established in planting boxes (52 cm long × 35 cm wide × 30 cm high) that contained a soil mixture comprising 18.5 kg of field soil from Shandong University Garden and 13.5 kg of peat medium (Pindstrup, Denmark).Coarse litter and stones were first removed from the field soil with a 5 mm sieve.In March 2021, seeds were planted at a depth of 2.5 mm following the experimental design described below.Each plant community was sown at a rate to produce ~1000 total sprouting seedlings, based on background germination rates measured prior to starting the experiment (Supplementary Methods, Table S1).Communities were planted with two levels of plant diversity (high, low).The high diversity treatment included all eight plant species.The low diversity treatment included four plant species, with two different plant communities used to control for species selection effects.The first low diversity community comprised one randomly selected plant species from each of the four families, with the second low diversity community comprised of the remaining plant species, such that one community included E. indica, Su. glauca, C. bonii, and G. soja, while the other community included Se. viridis, Su. salsa, X. strumarium and M. officinalis.This design resulted in three types of plant communities, with two levels of plant diversity.One month after seedling germination, each type of plant community was crossed with 11 global change treatments: nitrogen deposition, soil salinity, drought, plant invasion, grazing, oil pollution, plastics pollution, antibiotics pollution, heavy metal pollution, pesticide pollution, and a control community where no treatments were applied.Details of the protocol for each global change treatment are described in the Supplementary Methods.
In total there were 132 plant communities (3 plant community types × 11 global change treatments × 4 replicates).Each community received 4 L of water every 2 days except for those exposed to the drought treatment.

| Measurement of ecosystem multifunctionality
To assess ecosystem multifunctionality, we first measured nine separate ecosystem functions in each community, including primary productivity, microbial biomass, nitrogen stocks, phosphorus stocks, carbon stocks, floral abundance, litter decomposition, crown interception, and soil enzyme activity, after 18 weeks of growth under global change treatments.Measurement protocols for each ecosystem function are described in the Supplementary Methods.
Where F is the total number of ecosystem functions that were measured; f i is the measurement of a single ecosystem function i; r is the mathematical function to transform ecosystem function i to a positive value (see below); g is the transformation to standardize the ecosystem function i.Data for ecosystem functions for which lower values reflect a more desirable aspect of the ecosystem (i.e., lower soil total nitrogen, nitrate, ammonium, and phosphorus content) were multiplied by −1 (inverted around the 0 mean) to maintain directional change with other ecosystem functions, such that a decline from their desirable state corresponds to increasingly negative values (Wagg et al., 2014).This transformation allowed any general differences in multifunctionality among communities to be more easily assessed.All ecosystem function data were standardized by maximum transformation (all values in [0,1]), and these standardized data were then used to calculate the ecosystem multifunctionality metric.

| Measurement of complementarity and selection effects
Complementarity effects occur when, due to resource partitioning and positive interactions, the net yield (i.e., biomass) in high diversity plant communities exceeds that predicted based on yield measured in low diversity plant communities.Selection effects occur when species with higher-than-average yields in low diversity plant communities dominate the biomass of high diversity plant communities.
To calculate complementarity and selection effects in our experiment, we used the additive partitioning method, following the approach of Loreau and Hector (2001): Specifically, we assessed complementarity and selection effects between two kinds of low diversity plant communities and a high diversity plant community: For the calculation process, we first calculated the diversity effect (ΔY) in high diversity plant communities, which was the sum of the complementarity effect and selection effect.M low1 and M low2 represent the aboveground biomass of two kinds of low diversity plant community; M high1 and M high2 represent the aboveground biomass of the four plant species in the high diversity plant community that correspond to the two kinds of low diversity plant community; N represents the number of types of low diversity plant communities (a constant of 2 for this experiment).A positive diversity effect meant that the observed yield in high diversity plant communities (Y 0 ) exceeded the "expected yield" (Y E ) (i.e., ΔY > 0).Expected yield was calculated as the weighted (by the proportion of the total estimated number of seedlings for each plant species in high diversity plant communities, Table S1) average of the yields in low diversity plant communities.
Complementarity effects were calculated by multiplying the number of community types in low plant diversity communities (N), the average deviation from expected yield of the plants in the high diversity plant communities (ΔRY), and the average yield of each type of low diversity plant community (M).A positive complementarity effect meant that the yield of plants in the high diversity plant community was higher than expected based on the weighted average yield in the corresponding low diversity plant community.Selection effects were estimated as the covariance between the yield of community types in low diversity plant communities and the change in relative yield in high diversity plant communities [i.e., Ncov(ΔRY, M)], which could also be calculated as the difference between diversity effects and complementarity effects (i.e., N RYM − ΔY).A positive selection effect meant that species with higher-than-average yields in low diversity plant communities dominated the biomass of high diversity plant communities.

| Measurement of niche width and overlap
Niche width was used to describe the use of environmental resources by each plant community, while niche overlap refers to the degree of overlap in resource use between two or more species in the community (Bolnick et al., 2002).Because initial soil nutrients, growth space and greenhouse conditions were consistent, we attributed observed changes in niche characteristics (i.e., niche width and niche overlap) of plant communities to plant diversity and global change treatments.
First, we calculated the importance value (IV i ) for each species among all plant communities using the below formula (Curtis & McIntosh, 1951;Mori et al., 1983): AR i is the relative abundance of plant species i with respect to total abundance across all plant species; DR i is the relative cover (percent cover) of plant species i with respect to the sum of mean cover of all plant species; FR i was the relative frequency of species i with respect to the sum of frequencies of all plant species.
Second, we estimated niche width for any one species i and niche overlap for any two species i and k using the importance values of each plant species in each plant community at harvest, and based on Levin's niche width index (Levins, 1968) and the Schoener niche overlap index (Schoener, 1968), implemented with the "spaa" package (Zhang, 2016) in R (v 4.0.1)(R core team, 2023): P ij and P kj represent the importance value of species i and k in each plant community j.Finally, the total niche width for each plant community was calculated as the sum of Levin's niche width index of plant species i times the relative abundance of plant species i.Total niche overlap for each plant community was calculated as the mean Schoener niche overlap index of plant species i and k.

| Fungal community
To quantify the effects of global change factors and plant community diversity on the relative abundance of fungal plant pathogens, 25 g of soil was collected at harvest from 5 to 10 cm below the soil surface using a spade (sterilized in 75% ethanol) and following the 5-point sampling method (five individual 5 g samples collected from points in a quincunx [i.e., at the community center and halfway towards each of the four corners] and then pooled together for analysis).Soil samples were stored at −80°C until DNA extraction for fungal community analysis by high-throughput sequencing.The DNA extraction, library construction, and bioinformatic analysis protocols are described in the Supplementary Methods.Fungal plant pathogens were identified from the fungal sequencing data by using a consistent approach to query the FUNGuild database of fungal functional guilds (Nguyen et al., 2016): Any taxa assigned as "probable" or "highly probable" plant pathogens were used for subsequent analyses, including taxa that also belonged to other guilds.

| Measurement of response index to global change factors
To explore the impacts of global change factors on ecosystem multifunctionality and other variables that were measured in our experiment, we calculated a response index that standardized effects across diversity treatments: where X represented measurements of ecosystem multifunctionality, niche width, niche overlap, the abundance of novel functional plants Schoener niche overlap index for species i and and their interaction as explanatory variables (Table S3).To account for the non-independence of low diversity plant communities of the same community type, we included community type (e.g., low diversity plant community 1, low diversity plant community 2, high diversity plant community) as a random effect in the linear mixed model (Table S3).S3).
If the absolute value of the response index for one parameter was significantly higher for high than low diversity plant communities, we interpreted this as meaning that this parameter had greater sensitivity or reduced resistance to global change factors in high than low diversity plant communities (Figure 1).

| Mechanisms of diversity effects under global change factors
We assessed potential mechanisms that mediate the impact of plant This analysis was conducted using the "multifunc" package (Byrnes et al., 2014(Byrnes et al., , 2023)).Any plant species that differed between the two minimally adequate sets of species with and without global change factors were defined as novel functional plants that played an important role in ecosystem multifunctionality (Byrnes et al., 2014).
To test whether high and low diversity plant communities differed in how average and individual effects of global change factors impacted the response indexes of niche width, niche overlap, abundance of novel functional plants, and relative abundance of fungal plant pathogens, we again used a series of linear mixed models with response index as the response variable, plant community diversity as the explanatory variable, and community type as a random effect (Table S3).Post hoc Tukey tests with false discovery rate (FDR) correction were used to compare response index between high and low plant diversity communities.Kolmogorov-Smirnov and Levene tests were conducted to assess normality and homogeneity of residuals for all models, respectively, and these assumptions were satisfied.

| Relationships between ecosystem multifunctionality and plant diversity with global change factors
To test hypothesized relationships among plant diversity, global change factors, and ecosystem multifunctionality (Figure S2), we used a piecewise structural equation model (SEM) (Shipley, 2013).

| Global change factors reduced selection but not complementarity effects in plant communities
Selection effects were 60% lower in plant communities subjected with both complementarity effects (R 2 = 0.144, p = .006,Figure 3c) and selection effects (R 2 = 0.144, p = .006,Figure 3d).

| Global change factors affected niche width and overlap, abundance of novel functional plants and fungal pathogen relative abundance
The    & Callaway, 1994;He et al., 2013;i.e., increased complementarity effects).Second, the probability of high functioning species being present in a plant community increases with higher diversity (i.e., increased selection effects) under global change (Baert et al., 2018).
However, we found that the response index of niche overlap to global change factors did not differ between high and low diversity plant communities, suggesting that plant competition might not be disproportionately alleviated in high diversity plant communities.
Moreover, in our experiment, Chenopodiaceae species (Su.glauca available to lose but may remain higher than the corresponding low diversity plant communities.Therefore, the use of only two levels of plant diversity may limit our ability to draw more general conclusions (Tilman et al., 2006), and future studies could further explore these results across gradients of species diversity or in other study systems.
Global change factors consistently reduced selection effects, meaning that they reduced the relative abundance of species with disproportionally high contributions to ecosystem multifunctionality of plant communities (Dieleman et al., 2015;Loreau & Hector, 2001;Lyons et al., 2020;Tilman et al., 2014).We suggest that this reduction in selection effects might contribute to why ecosystem multifunctionality of high diversity plant communities was gen cycling, primary productivity and other ecosystem functions (Drinkwater et al., 2021;Taylor et al., 2020), and previous studies have shown that legumes and their symbiosis with nitrogen-fixing rhizobia bacteria are sensitive to global change (Newton et al., 2014;Reverchon et al., 2012;Scheelbeek et al., 2018).Moreover, this result was consistent with those of a species richness experiment by global change is only predicted to worsen in the future (Barreto & Lindo, 2022;Rillig et al., 2019;Song et al., 2019;Speißer et al., 2022).van der Plas, 2019).Our results demonstrate that ecosystem multifunctionality has greater provisioning and less stability in response to global change in high diversity plant communities, which also support multiple ecosystem functions (Hong et al., 2022;Soliveres et al., 2016; Figure S3), further highlighting the importance of protecting high biodiversity ecosystems (Xu et al., 2017).Therefore, we suggest that researchers and practitioners interested in the

| CON CLUS ION
cies) and high (eight species) diversity plant communities were subjected to 10 global change factors (nitrogen deposition, soil salinity, drought, plant invasion, simulated grazing, oil pollution, plastics pollution, antibiotics pollution, heavy metal pollution, and pesticide pollution) and the responses of nine ecosystem functions (primary productivity, microbial biomass, nitrogen stocks, phosphorus stocks, carbon stocks, floral abundance, litter decomposition, crown interception, and soil enzyme activity) were measured to quantify ecosystem multifunctionality.To investigate potential mechanisms through which plant community diversity interacts with global change factors to influence ecosystem multifunctionality, we used plant aboveground biomass data to quantify changes in complementarity and selection effects [based on the additive partitioning method (Loreau & Hector, 2001)], and niche width and overlap [based on Levin's | 3 of 14 XU et al.
, respectively] in response to global change factors.Specifically, we investigated: 1.Whether high diversity plant communities had higher absolute ecosystem multifunctionality and a weaker response to average and individual effects of global change factors compared with low diversity plant communities.2. Whether global change factors affected the relationship between plant diversity and ecosystem multifunctionality via changes in complementarity and/or selection effects.We predicted that if complementarity effects were important, the impacts of multiple global change factors would be related to changes in plant niche width and overlap.Alternatively, if selection effects were important, the impacts of global change factors would be related to changes in the abundance of novel functional plants and relative abundance of fungal plant pathogens.
Gramineae [Eleusine indica (L.) Gaertn and Setaria viridis (L.) Beauv.];Chenopodiaceae [Suaeda glauca Bunge and Suaeda salsa (Linnaeus) Pallas]; Compositae (Crepis bonii Gagnepain and Xanthium strumarium L.); and Fabaceae [Glycine soja Sieb.et Zucc.and Melilotus officinalis (L.) Pall.].We selected two plants per family because this allowed us to minimize genetic differences between the two types of low diversity plant community (see Section 2.2).Seeds of each F I G U R E 1 Hypothesized relationships between plant community diversity and ecosystem multifunctionality in response to global change factors.Under scenario (a), ecosystem multifunctionality is more resistant to global change factors in high compared with low diversity plant communities.Under scenario (b), ecosystem multifunctionality is more sensitive to global change factors in high compared with low diversity plant communities.Under scenario (c), there is no difference in the impact of global change factors on ecosystem multifunctionality between high and low diversity plant communities.Under scenario (d), there is no impact of global change factors on ecosystem multifunctionality in high and low diversity plant communities.Points represent the hypothesized mean response index (±SE) in high (orange circles) and low (blue triangles) diversity plant communities, respectively.+ = global change factors with a positive effect on ecosystem multifunctionality; − = global change factors with a negative effect on ecosystem multifunctionality; O = global change factors with no effect on ecosystem multifunctionality (dashed line).

(
see below), and the relative abundance of fungal plant pathogens.If the value of the response index was >0, this meant that global change Selection effect = Ncov(ΔRY, M) = ΔY − Complementarity effect ln X with global change factor − ln X without global change factor , factors had a positive effect on the variable of interest.If the response index was <0, this meant that global change factors had a negative effect.After calculation of the response index to each individual global change factor in low or high plant diversity communities, the response index data were averaged across all global change treatments to represent the average response index to global change factors for low and high diversity plant communities.2.8 | Statistical analysis 2.8.1 | Absolute value and response index of ecosystem multifunctionality to global change factors in high and low diversity plant communities To test the average effect of global change factors on the absolute value of ecosystem multifunctionality, we used a linear mixed model with the absolute value of ecosystem multifunctionality as the response variable and plant community diversity, global change treatment (i.e., presence or absence of any global change treatment) Post hoc Tukey tests with false discovery rate (FDR) correction were used to compare treatment means among the four global change and plant diversity treatment combinations.To explore the differences of absolute value of ecosystem multifunctionality with or without individual global change factors, a series of one-sample t-tests with FDR correction were used.To test whether the response index for average and individual global change factors was different from 0, we again used a series of one-sample t-tests with FDR correction.To test whether the impact of average and individual global change factors on ecosystem multifunctionality differed between high and low diversity plant communities, we used multiple linear mixed models (e.g., response index for ecosystem multifunctionality ~ plant community diversity) with FDR correction.To account for the non-independence of low diversity plant communities of the same community type, we included community type as a random effect in the linear mixed model (Table diversity on the response of ecosystem multifunctionality to global change factors, including complementarity and selection effects.To test whether complementarity and selection effects differed between communities that were affected or unaffected by average and individual effects of global change factors, we again used a linear mixed model with complementarity and selection effects as response variables, global change treatments as explanatory variables, and community type as a random effect (TableS3).Post hoc Tukey tests with false discovery rate (FDR) correction were used to compare treatment means between control and global change treatments.To test the relationships of ecosystem multifunctionality with complementarity effects and selection effects in the high diversity plant community, we used simple linear regression.To identify novel functional plants, we obtained the minimally adequate set of species that impacted ecosystem multifunctionality with or without global change factors by using a stepwise AIC (Akaike Information Criterion) model selection approach to fit linear additive models to ecosystem multifunctionality, followingHector and Bagchi (2007).Two initial full models explaining ecosystem multifunctionality with or without global change factors contained main effects for all plant species.Each species was removed from the models in turn to obtain multiple models, for which AIC was calculated.The AIC values of the resulting models were compared, and the plant species whose exclusion led to the greatest improvement (reduction) in the AIC value was permanently removed from the model.This model selection process was repeated until dropping any of the remaining species increased AIC, at which point we considered that the minimally adequate model had been identified.

| 7 of 14 XU|
et al.In the SEM, communities subjected to global change factors were represented by 1, and communities without global change factors (i.e., control communities) were represented by 0 (Wu et al., 2022).Because the abundance of novel functional plants and the relative abundance of fungal plant pathogens were identified as contributing to the resistance of ecosystem multifunctionality to global change factors, these variables were also included in the SEM.We therefore hypothesized that global change factors and plant diversity would directly influence ecosystem multifunctionality as well as having indirect effects via changes in the abundance of novel functional plants and the relative abundance of fungal plant pathogens.Nonsignificant relationships in the theoretical model (Figure S2) were removed to optimize model fit, which was presumed adequate with a non-significant χ 2 test (p > .05),low root means square error of approximation (RMSEA <0.08), and high comparative fit index (CFI >0.90) in AMOS version 24.0 (Amos Development, Spring House, Ecosystem multifunctionality was more sensitive to global change factors in high diversity compared with low diversity plant communities Plant diversity and global change factors affected ecosystem multifunctionality of plant communities.Ecosystem multifunctionality was 18% and 23% lower on average in low and high diversity plant communities, respectively, that were subjected to global change factors compared with those where global change factors were absent (control communities) (Figure 2a).Moreover, ecosystem multifunctionality did not differ between control communities with low and high plant diversity, but was 8% lower in low compared with high diversity plant communities under the average effects of global change (Figure 2a).Ecosystem multifunctionality was 30% more sensitive (i.e., a 30% more negative response index) to the average effect of global change factors in high diversity than in low diversity plant communities (F = 19.21,df = 1, p < .001,top row of Figure 2b).Results for individual global change factors are presented in the Supplementary Results and Figure 2.
Figure 3b).In contrast, the average effect of global change factors did not significantly alter complementarity effects (F = 1.67, df = 1, p = .203,top two rows of Figure 3a).Results for individual global change factors are presented in the Supplementary Results andFigure 3. Ecosystem multifunctionality had a positive relationship Figure 4c).Results for individual global change factors are presented in the Supplementary Results and Figure 4c.On average, global change factors had a positive impact on the relative abundance of fungal plant pathogens, but the average response was only significantly different from zero in low diversity plant communities (t = 4.65, df = 1, p < .001,top row of Figure 4d).Moreover, low and high diversity plant communities did not differ in their response index of fungal pathogen relative abundance to the average effect of global change factors (F = 0.94, df = 1, p = .332,top row of Figure 4d).Results for individual global change factors are presented in the Supplementary Results and the other rows of Figure 4d.

3. 4 |
Novel functional plants and fungal pathogens mediated the impacts of global change factors and plant diversity on ecosystem multifunctionalityThe SEM supported that plant community diversity (positive) and global change factors (negative) had opposite direct relationships with ecosystem multifunctionality, although the negative relationship with global change factors was more than twice as strong (Figure5).Plant community diversity also had a positive indirect relationship with ecosystem multifunctionality, mediated by its positive relationship with the abundance of novel functional plants, which in turn had a strong positive relationship with ecosystem multifunctionality.In contrast, global change factors had a negative indirect relationship with ecosystem multifunctionality that was mediated by its negative direct relationship with the abundance of novel functional plants.Moreover, global change factors also had a positive direct relationship with the relative abundance of fungal plant pathogens, which in turn had a negative direct relationship with the abundance of novel functional plants, and therefore an indirect negative relationship with ecosystem multifunctionality (Figure5).Finally, the negative linear relationship between the response indexes of fungal pathogen relative abundance and the abundance of novel functional plants was only observed in high diversity plant communities (FigureS1).

4
| DISCUSS ION Contrary to prevailing theory, our experiment showed that ecosystem multifunctionality was more sensitive to the impacts of global change factors in high diversity compared with low diversity plant communities, although global change treatments decreased almost all individual ecosystem functions (Figure S4).This result suggests that high plant diversity may not mitigate negative impacts of global change factors on ecosystem multifunctionality, which F I G U R E 2 Mean (±SE) absolute values of ecosystem multifunctionality (a) and its response index to the average or individual effects of global change factors (b) in high (orange circles) and low (blue triangles) diversity plant communities.In the grey portion of panel A (upper two rows), different lowercase letters indicate significant differences in ecosystem multifunctionality between high and low diversity plant communities with or without the average effect of ten global change factors (based on the results of a linear mixed model and post hoc Tukey analysis with false discovery rate [FDR] correction).The lower part of panel (a) shows the results from linear mixed models with FDR correction that were used to test whether the absolute value of ecosystem multifunctionality differed between communities subjected to a single global change factor and their respective control community: ***p < .001;**.001 < p < .01;*.01 < p < .05;ns p > .05.In panel (b), green and purple lines represent response index values that were stronger or weaker, respectively, in high compared with low diversity plant communities.The dashed vertical line represents the null expectation of the response index of ecosystem multifunctionality as 0; that is, ecosystem multifunctionality was not affected by the global change factor(s).All global change factors reduced ecosystem multifunctionality of both high and low diversity plant communities, based on one-sample t-tests (all p < .026)with false discovery rate (FDR) correction.Linear mixed models with FDR correction were used to test whether each response index differed between low and high diversity plant communities: ***p < .001;**.001 < p < .01;*.01 < p < .05;ns p > .05. we demonstrate might result from reduced selection effects via the decreased abundance of novel functional plants (i.e., legumes, Fabaceae: Glycine soja and Melilotus officinalis).Moreover, global change factors inhibited the abundance of novel functional plants by promoting fungal plant pathogens, which might contribute to the sensitivity of ecosystem multifunctionality to global change factors in high diversity plant communities.Our study examined the response of ecosystem multifunctionality to multiple global change factors and investigated potential mechanisms underlying differences in resistance of ecosystem multifunctionality between low and high diversity plant communities.Below we discuss the implications of our results for the understanding of how plant communities respond to complex global change.It is generally thought that high diversity communities are more resistant to environmental changes(Garcia et al., 2018) and that ecosystem functioning in high diversity communities is less impacted by global change (Hong et al., 2022).Contrary to theory and prevailing evidence, we found that ecosystem multifunctionality was more sensitive to global change factors in high diversity than in low diversity plant communities.Previous studies have identified two possible explanations for why ecosystem functioning should be more resistant to global change factors in high compared to low diversity communities.First, in stressful environments (i.e., under global change), interspecific competition could be reduced or even switch to facilitation in high diversity plant communities (Bertness

F
Mean (±SE) complementarity effect index (a) and selection effect index (b) in response to the average or individual effects of multiple global change factors, based on the additive partitioning method(Loreau & Hector, 2001), and their linear relationships with ecosystem multifunctionality in high diversity plant communities (c, d).Complementarity effects occurred when the net yield (i.e., biomass) in high diversity plant communities exceeded that predicted based on the yield in low diversity plant communities, due to resource partitioning and positive interactions.Selection effects occurred when species with higher-than-average yields in low diversity plant communities dominated the high diversity plant communities.Panels (a) and (b) show the results of linear mixed models with false discovery rate (FDR) correction that were used to test whether the mean effect index differed between control communities that were unaffected by global change factors (green points) and communities where individual global change factor treatments were applied (purple points): ***p < .001;**.001 < p < .01;*.01 < p < .05;ns p > .05.The lack of symbols in panel (a) indicates that no global change factors affected the complementarity effect index relative to the control treatment (all p > .05).Panels (c) and (d) show the relationships between ecosystem multifunctionality and the complementarity effect index and selection effect index.and Su.salsa) were dominant across high and low diversity plant communities that were both affected and unaffected by global change factors, indicating that high functioning species were present in all communities.Therefore, in our experiment, changes in plant competition and dominant species were inconsistent with predictions based on previous studies, which may have contributed to the high sensitivity of ecosystem multifunctionality to global change factors in high diversity plant communities.However, an alternative explanation may emerge based on the high absolute values of ecosystem multifunctionality observed in high diversity plant communities under global change (Figure S3).For example, if ecosystem multifunctionality was already low in low diversity plant communities, there is little scope for it become much lower, hence the relatively small response index.Conversely, ecosystem multifunctionality of high diversity plant communities might be more strongly impacted by global change because it has more ecosystem multifunctionality

F
I G U R E 4 Mean (±SE) response index of niche width (a), niche overlap (b), the abundance of novel functional plants (c), and the relative abundance of fungal plant pathogens (d) in high (orange circles) and low (blue triangles) diversity plant communities in response to the average or individual effects of multiple global change factors.Green and purple lines represent a response index that was stronger or weaker, respectively, in high compared with low diversity plant communities.Linear mixed models with FDR correction tested whether each response index differed between low and high diversity plant communities: ***p < .001;**.001 < p < .01;*.01 < p < .05;ns p > .05.The dashed vertical line represents the null expectation of the response index of ecosystem multifunctionality as 0. Points with a cross (×) in the middle show where global change factors had no significant effect on the relative ecosystem function, whereas points without a cross show where global change factors had a significant effect (p < .05),based on one-sample t-tests with false discovery rate (FDR) correction.F I G U R E 5 Structural equation model (SEM) relating plant diversity, global change factors, and ecosystem multifunctionality.Green boxes represent experimental treatments (plant diversity, global change factors), blue boxes represent response variables (ecosystem multifunctionality), and yellow boxes represent observed explanatory variables (abundance of novel functional plants, relative abundance of fungal plant pathogens).Blue lines represent positive relationships, red lines represent negative relationships, and line thickness depicts the magnitude of path coefficients (shown beside line, with statistical support).The fitting values of the model were χ 2 = 4.539, p = .209,root mean square error of approximation (RMSEA) = 0.063, and comparative fit index (CFI) = 0.988.| 11 of 14 XU et al.more responsive to the impacts of global change factors(Godoy et al., 2020).Although selection effects were quantified using aboveground biomass (which might only reflect the ecosystem function of primary productivity), we found a significant positive relationship between selection effects and ecosystem multifunctionality, indicating that selection effects might mediate the effects of diversity and global change factors on multiple ecosystem functions, rather than only primary productivity(Loreau & Hector, 2001;Searle & Chen, 2020;Tilman et al., 2014; Figure  3d).This explanation was further supported by the stronger negative impacts of global change factors on the abundance of novel functional plants (identified as the leguminous plants G. soja and M. officinalis) in high compared with low diversity plant communities.Legumes promote nitro- Jia et al. (2020), who showed that legumes contributed mostly to selection effects, which in turn influenced positive diversity effects on ecosystem functioning.Therefore, the more sensitive response of novel functional plants (rather than of dominant species) in high compared with low diversity plant communities might be a contributing mechanism behind reduced selection effects and ecosystem multifunctionality under global change.We present evidence that fungal plant pathogens also contributed to the impacts of global change factors on ecosystem multifunctionality.Global change factors promoted the relative abundance of fungal plant pathogens in our experiment, potentially through inducing physiological stress on host plants and decreasing their resistance to fungal pathogens(Liu et al., 2019;McElrone et al., 2003).Moreover, we also observed a negative relationship between the response index of relative abundance of fungal plant pathogens and the response index of the abundance of novel functional plants (i.e., legumes), but only in high diversity plant communities.We interpret this result as indicating that the negative relationship between fungal pathogens and novel functional plants was strengthened by increased plant diversity.Novel functional plants disproportionately contributed to ecosystem multifunctionality, and therefore, through impacting novel functional plants, fungal pathogens also indirectly contributed to the impacts of plant diversity and global change factors on ecosystem multifunctionality.Our results emphasize the importance of interactions between the plant community and belowground fungal taxa on ecosystem multifunctionality under global change.Although ecosystem multifunctionality of high diversity plant communities was highly sensitive to the average impact of multiple global change factors, the responses of ecosystem multifunctionality to individual global change factors were variable.However, ecosystem multifunctionality was more sensitive to most global change factors (except for nitrogen deposition) in high compared with low diversity communities, including soil salinity, drought, plant invasion, oil pollution, and pesticide pollution.The contradictory resultfor nitrogen deposition might be explained by nitrogen deposition also having a more negative impact on niche width in high compared with low diversity plant communities.Besides the mediated effects of plant diversity on the ecosystem multifunctionality, some global change treatments might directly affect ecosystem functioning, such as nitrogen deposition increasing the soil nitrogen pool (FigureS3).Taken together, these mixed effects indicate that the impacts of global change factors on ecosystem multifunctionality are complex(Eisenhauer et al., 2018) and that interactions between multiple global change factors might counteract or intensify the effects of single global change factors on ecosystem multifunctionality in natural ecosystems.Therefore, studies that investigate the interactive effects of multiple global change factors on ecosystem multifunctionality will be an important future research direction, especially as Although impacts varied among individual global change factors, ecosystem multifunctionality was more sensitive to impacts from global change factors in high diversity compared with low diversity plant communities.We show that this higher sensitivity might be due to weakened selection effects, as global change factors reduced the abundance of novel functional plants, such as legumes, in high diversity plant communities.Moreover, fungal plant pathogens also mediated the negative effects of global change factors on ecosystem multifunctionality, potentially through their inhibition of legumes in high diversity plant communities.The increasing complexity of global change threatens the planet with biodiversity loss, posing a dire risk to ecosystem functioning(Oliver et al., 2015; management of ecosystem multifunctionality under global change should prioritize the protection of high biodiversity ecosystems and sensitive but novel functional plants, such as legumes, as well as monitoring and managing fungal plant pathogens.Conceptualization; data curation; formal analysis; investigation; methodology; project administration; resources; software; supervision; validation; visualization; writing -original draft; writing -review and editing.Warwick J. Allen: Conceptualization; formal analysis; validation; visualization; writing -review and editing.Xiaona Yu: Conceptualization; methodology; validation; writing -original draft.Yi Hu: Data curation; resources; visualization.Jingfeng Wang: Conceptualization; data curation; resources;