Resource stoichiometry shapes community invasion resistance via productivity-mediated species identity effects

Abstract

Diversity–invasion resistance relationships are often variable and sensitive to environmental conditions such as resource availability. Resource stoichiometry, the relative concentration of different elements in the environment, has been shown to have strong effects on the physiology and interactions between different species. Yet, its role for diversity–invasion resistance relationships is still poorly understood. Here, we explored how the ratio of nitrogen (N) and phosphorus affects the productivity and invasion resistance of constructed microbial communities by a plant pathogenic bacterium, Ralstonia solanacearum. We found that resource stoichiometry and species identity effects affected the invasion resistance of communities. Both high N concentration and resident community diversity constrained invasions, and two resident species, in particular, had strong negative effects on the relative density of the invader and the resident community productivity. While resource stoichiometry did not affect the mean productivity of the resident community, it favoured the growth of two species that strongly constrained invasions turning the slope of productivity–invasion resistance relationship more negative. Together our findings suggest that alterations in resource stoichiometry can change the community resistance to invasions by having disproportionate effects on species growth, potentially explaining changes in microbial community composition under eutrophication.

1. Introduction

Microbial biodiversity plays an important role in ecosystem functioning by offering sets of functions that cannot be provided by single species [1–3]. For example, host-associated microbial communities can ward off pathogens, thereby protecting their associated host organism [4–6]. This process can also be viewed from the perspective of biological invasions where the members of resident microbial communities facilitate or constrain the establishment of the invader [7,8]. Several studies have shown that increasing community diversity reduces the likelihood of invasions by promoting a more comprehensive use of available niches in the given environment [3,9,10]. Such diversity–invasion resistance relationships are, however, often sensitive to environmental conditions such as resource availability [11] and temperature [12]. As a result, we still poorly understand how environmental contexts shape diversity–invasion resistance relationships.

Resource availability has been shown to be an important factor affecting the outcome of biological invasions [13–15]. Mechanistically, concentration or composition of resources can alter the physiology and interactions between different species within communities which can then lead to changes in community invasion resistance. Moreover, changes in resource availability may change the significance of species identity effects, i.e. the contribution of resident community members to the invasion, by promoting the growth of species that grow slow or fast [16,17]. Species identity effects could, thus, explain positive diversity–invasion resistance relationships across environmental gradients where different species contribute to the invasion resistance under specific environmental conditions [1,18]. Here, we studied how resource stoichiometry, the relative concentration of different elements in the environment, shapes invasions via diversity, productivity and species identity effects.

Resource stoichiometry is a broad and active research field in ecology that has been extensively used to understand predator–prey interactions [19,20]. While several studies have highlighted the importance of resource stoichiometry for the ecology and functioning of communities [21,22], its effects have been less studied in the context of diversity–ecosystem functioning relationships. Environmental stoichiometry can be used to link tissue composition of organisms with trophic-level interactions [23–25] and it plays an important role in determining which species are able to grow in any given ecosystem affecting consumer–resource interactions [26,27]. For example, low carbon : phosphorus (C : P) ratio has been shown to favour fast-growing species leading to an increase in microbial diversity [28]. Resource stoichiometry can also affect invasions by altering species biomasses and growth dynamics [29]. However, it is unclear how resource stoichiometry shapes diversity–invasion resistance and productivity–invasion resistance relationships.

In the present study, we used an experimental approach to directly expose model microbial communities with varying levels of diversity to invasions by a single invader species under different resource stoichiometry treatments. The model ‘resident’ community was constructed by using five different bacterial species that have previously been shown to constrain invasions in a diversity-dependent manner [4,17]. We used a plant pathogenic Ralstonia solanacearum bacterium as an invader whose life cycle is directly linked to biological invasions of the plant rhizosphere microbiome. Ralstonia solanacearum causes bacterial wilt disease [30,31] and is a major threat to global food production [32]. Before infecting its host, R. solanacearum must first get through microbial communities surrounding the plant roots. It has previously been shown that competition for resources between the invader and resident community members is important for the outcome of invasions [4,17]. How these invasion outcomes are affected by dynamic changes in nutrient levels typical for rhizosphere microbiomes [33,34], and resident community diversity remains unclear. To study this, we manipulated both resident community diversity gradient (richness levels of 1 to 5 species in all possible combinations) and the resource stoichiometry of the environment by changing the relative concentration and ratio of nitrogen (N) and phosphorus (P) orthogonally by following the Redfield ratio. The Redfield ratio is the atomic ratio of carbon (C), N and P found in phytoplankton and throughout the deep oceans [35] and a general baseline of element composition for aquatic and terrestrial ecosystems [36,37]. Communities were then exposed to R. solanacearum invasions and the invasion success was determined as the relative density of R. solanacearum invader after 72 h growth in the resident community (indicative of the reproductive success of the invader): the higher the final relative abundance of R. solanacearum, the higher the invasion success. We expected that resident community diversity–invasion resistance relationship could be sensitive to resource stoichiometry having positive or negative effects on invasions, depending on specific changes in species' ability to grow under different N : P ratios. Mechanistically, changes in invasion outcomes could potentially be explained via effects on community productivity or changes in the relative contribution of community members to invasions via species identity effects.

2. Methods

(a). Bacterial strains and plasmids

We used R. solanacearum species QL-Rs1115 tagged with pYC12-mCherry plasmid as a model invader in our experiments [38]. Five avirulent, but closely related, Ralstonia spp. isolates (Ralstonia mannitolilytica QL-A2, Ralstonia mannitolilytica QL-A3, Ralstonia pickettii QL-A6, Ralstonia taiwanensis QL-117 and Ralstonia pickettii QL-140) were used to construct our model resident communities [4]. None of these bacteria showed direct antagonism towards each other or the invader, which suggests that they probably interact indirectly through competition for shared resources. A more detailed description of the bacteria and used plasmid can be found in the electronic supplementary material, table S1. All bacteria were stored at −80°C in 20% glycerol prior to the experiments.

(b). Assembly of resident communities

The resident communities were assembled by using all five avirulent species in substitutive design so that the final communities covered all possible species combinations and richness levels (total of 31 communities with equal initial bacterial biomasses, electronic supplementary material, table S2). Prior to the experiments, bacteria were pre-cultured from frozen stocks on nutrient agar (NA) plates (glucose 10.0 g l⁻¹, tryptone 5.0 g l⁻¹, beef extract 3.0 g l⁻¹, yeast extract 0.5 g l⁻¹, agar 15.0 g l⁻¹, pH 7.0) and single colonies were picked and re-grown in liquid nutrient broth (NA medium without agar) at 30°C for 12 h with 170 r.p.m. agitation. Bacterial isolates were washed three times in 0.85% NaCl to remove nutrient residues and re-suspended in 0.85% NaCl with final densities of 10⁷ cells ml⁻¹.

(c). Manipulation of resource stoichiometry

To manipulate the resource stoichiometry, we first set up a minimal salt medium, which did not contain C, N or P (MOPS 30 mM, CaCl₂ 0.1 mM, FeSO₄ 3 mM, KCl 20 mM, MgCl₂ 2 mM, Na₂SO₄ 14 mM and NaCl 51 mM, pH 7.0). The minimal medium was then supplemented with a mixture of carbons (fructose, glucose, sucrose, maltose, arabinose and galactose) in equal concentrations to yield a total concentration of 10 mM for all combined C resources, as described previously [4]. The concentration of total C resources (10 mM) was held constant for all resource stoichiometry treatments. To manipulate the concentration and ratio of N and P, we added NH₄Cl or NaH₂PO₄·2H₂O as the sole N and P resource, respectively. A total of six resource stoichiometry treatments with four unique N : P ratios were established for the experiment where low, intermediate and high N levels were established within both low and high P levels (table 1). Each medium was then used to establish replicate treatments on 96-well microtiter plates in triplicate (18 microplates in total) for each resident community combination.

Table 1.

Concentration of nitrogen (NH₄Cl) and phosphorus (NaH₂PO₄) and their ratios (N : P) in different treatments.

treatment	nitrogen (mM)	phosphorus (mM)	N : P ratio
1	1.5	0.09	16 : 1 (Redfield ratio)
2	15	0.09	160 : 1
3	0.15	0.09	1.6 : 1
4	1.5	0.9	1.6 : 1
5	15	0.9	16 : 1
6	0.15	0.9	1.6 : 10

(d). Measuring resident community invasion resistance and productivity in microcosms

To quantify invasion resistance, all communities (10⁶ cells ml⁻¹ in 200 µl of final volume) were exposed to invasion by R. solanacearum QL-Rs1115 (10⁵ cells ml⁻¹ in 200 µl of final volume) under different resource stoichiometry environments. Replicate communities without invader were used as control treatments. All communities were incubated for 72 h at 30°C with 170 r.p.m. orbital agitation. To measure invader density relative to resident community density, we measured the mCherry fluorescence signal (excitation: 587 nm, emission: 610 nm, gain: 60) of the invader and calculated the invasion success as mCherry relative fluorescence unit against total bacterial density of the community (RFU, mCherry/OD₆₀₀) at the end of the experiment (after 72 h of incubation). To quantify the total productivity of different communities in each resource environment, we used optical density (OD₆₀₀) as a measure of total bacterial growth (invader and the resident community). We used the control communities without the invader to blank the fluorescence signal background and optical density of the culture media to blank the OD₆₀₀ background.

To verify plasmid stability during the invasion experiments, we grew gentamycin-tagged mCherry plasmid carrying R. solanacearum invader (10⁶ cells ml⁻¹ in 200 µl of final volume) in four N : P ratios (0.16, 1.6, 16 and 160) in the absence and presence of gentamycin antibiotic (30 µg ml⁻¹) for 72 h. Gentamycin was added only at the beginning or at every 24 h to create a strong selective pressure on the plasmid. The plasmid stability was determined as fluorescent signal intensity, which is indicative of bacterial growth and expression of the plasmid-encoded mCherry fluorescent protein (electronic supplementary material, figure S1). No difference was observed between different antibiotic treatments in any of the N : P ratios after 72 h of incubation (electronic supplementary material, figure S1; The main effect of antibiotic treatment in 0.16, 1.6, 16 and 160 N : P ratios, respectively: F_1,194 = 0.08, p = 0.77; F_1,381 = 0.2, p = 0.65, F_1,371 = 1.57, p = 0.21 and F_1,190 = 0.08, p = 0.77). This suggests that the plasmid was stably maintained during the invasion experiments in the absence of gentamycin.

(e). Measuring the growth and consumption of nitrogen and phosphorus by each bacterial species

The growth rate and productivity of all bacterial species were measured in monoculture at four N : P ratios. Bacterial species were inoculated at an initial density of 10⁶ cells ml⁻¹ in 96-well microtiter plates, as described above. Each monoculture was grown in triplicate under each of the N : P ratios at 30°C with agitation (170 r.p.m.) for 72 h. To determine growth rates, we measured bacterial growth with a spectrophotometer (OD₆₀₀) every 8 h to fit in logistics model function (‘gcFitModel’ in package ‘grofit’ in R 3.3.1) [39] and the maximum slope (μ, h⁻¹) of the logistic model was considered as the maximum growth rate [40]. The productivity of each species was determined as the bacterial biomass after 72 h. Culture medium without bacteria was used to determine the background absorbance before determining growth rates and productivity. To test the consumption of N and P by all bacteria, we obtained cell-free supernatant by centrifugation (10 000 r.p.m. for 10 min) and filtration (0.22 µm filters) after 72 h incubation. N and P concentrations were measured using a continuous-flow analyzer (AA3, SEAL, Germany) and compared with unconsumed media (no bacterial inoculation).

(f). Statistical analyses

Invasion success (RFU, mCherry/OD₆₀₀) and resource stoichiometry ratios (N : P) were log₁₀–transformed before statistical analyses to fulfil model assumptions. All comparisons between the growth of individual species were analysed using ANOVA and linear regression. General linear mixed models were used to examine the resident species identity effects and resource stoichiometry as a function of invader relative density (invasion success). Model 1 (richness-ratio) and model 2 (identity-ratio) were used to identify how individual species contributed to the invasion success as a function of resource stoichiometry. Model 3 (‘richness-concentration’, electronic supplementary material, table S3) was used to analyse whether N and P concentrations had interactive effects with species richness in determining invasion success and resident community productivity. Model 4 (‘productivity-ratio’, electronic supplementary material, table S4) was used to study the interactive effects between productivity and resource stoichiometry on the invasion success, while models 5 (‘richness-ratio’, electronic supplementary material, table S4) and 6 (‘identity-ratio’, electronic supplementary material, table S4) were used to explore the effects of community diversity and species identity on the resident community productivity. In some cases, we used the means of community treatment replicates for simplified analysis (e.g. species presence effects in figure 2).

Figure 2.

The effect of R. pickettii QL-A6 and QL-140 species on the relative density of the invader (panel (a) and (b), respectively). Relative density of the invader was defined as log-10 transformed relative mCherry fluorescence unit (RFU, mCherry/OD₆₀₀) after 72 h incubation. The 0 and 1 on x-axes denote for the presence and absence of R. pickettii QL-A6 and QL-140 species in the bacterial community and bars show ± 1 standard error (n = 186). Asterisks indicate significant differences (*p < 0.05; **p < 0.01; ***p < 0.001).

A structural equation model (SEM; ‘lavaan’ package in R) was used to investigate the relative importance of resource stoichiometry, species identity effects, species maximum growth rates, N and P consumption and resident community productivity on the invasion success (the prior model is shown in the electronic supplementary material, figure S2). All analyses were performed with R 3.3.1 [41].

3. Results

(a). Effects of resource stoichiometry on the diversity–invasion relationship

We found that increasing resident species richness had a negative effect on invasions across all N : P ratios (figure 1; model 1 in table 2, the main effect of species richness on invader relative density). Similarly, the relative density of the invader decreased with increasing N : P ratio and N concentration (model 1 in table 2, the main effect of N : P ratio on invader relative density), while P concentration alone had no effect (model 3 in the electronic supplementary material, table S3). No interactive effect between species richness and N : P ratio on the relative density of the invader was found (model 1 in table 2). These results suggest that both species richness and N : P ratio constrained invasions independently, while N concentration alone had a stronger effect than P concentration.

Figure 1.

The effects of species richness and N : P ratio on invasion success (relative invader density). The relative density of the invader was defined as log-10 transformed relative mCherry fluorescence unit (RFU, mCherry/OD₆₀₀) after 72 h incubation.

Table 2.

ANOVA table summarizing the species richness, N : P ratio and species identity effects on the relative density of the invader (models 1–2). (Significant effects (p < 0.05) are highlighted in italics and the ‘up’ and ‘down’ arrows denote for positive and negative effects on the relative density of invader, respectively. Non-significant terms were not retained in the final models (not retained).)

	relative density of the invader
	d.f.	F	p
model 1 (richness-ratio)
species richness (richness)	1	28.50	<0.001 ↓
N : P ratio (ratio)	1	9.71	0.0021 ↓
richness * ratio	1	0.0072	0.93
residuals	182
model summary	AIC: 368.45, R2 = 0.17
model 2 (identity-ratio)
QL-A2	not retained
QL-A3	not retained
QL-A6	1	524.41	<0.001 ↓
QL-117	not retained
QL-140	1	25.27	<0.001 ↓
N : P ratio (ratio)	1	33.74	<0.001 ↓
QL-A2 * ratio	not retained
QL-A3 * ratio	not retained
QL-A6 * ratio	not retained
QL-117 * ratio	not retained
QL-140 * ratio	not retained
residuals	182
model summary	AIC: 136.72, R2 = 0.76

(b). The effect of resident species identities on resident community productivity and invasions

We found that R. mannitolilytica QL-A2, R. mannitolilytica QL-A3 and R. taiwanensis QL-117 resident species had no significant species identity effects on the relative density of the invader (model 2 in table 2). By contrast, R. pickettii QL-A6 and R. pickettii QL-140 resident species were very effective at reducing the relative density of the invader (model 2 in table 2). To examine this in more detail, we compared the invasion resistance of resident communities with and without these two species. As expected, resident communities were less resistant to invasions in the absence of these two species (figure 2, comparison on communities with and without R. pickettii QL-A6: panel (a), F_1,184 = 400.4, p < 0.001, comparison on communities with and without R. pickettii QL-140: panel (b), F_1,184 = 4.47, p = 0.036). Mechanistically, this could be explained by relatively more efficient consumption of N and P compared to the other resident species (electronic supplementary material, figures S3 and S4; Species main effects at 0.15, 1.5 and 15 mM N concentrations, respectively: F_6,35 = 254.6, p < 0.001; F_6,35 = 3196, p < 0.001 and F_6,35 = 55.69, p < 0.001. Species main effects at 0.09 and 0.9 mM P concentrations, respectively: F_6,56 = 20.92, p < 0.001 and F_6,56 = 6.00, p < 0.001). However, no clear differences were found in comparison with the invader, which suggests that R. pickettii QL-A6 and QL-140 were equally good at consuming N and P (p > 0.05 in all pairwise comparisons).

Even though increasing the N : P ratio constrained invasions, no significant interactions with species identities were found in the full-scale invasion experiment (model 2 in table 2). To explore this further, we compared the growth of resident species and the invader separately in monocultures. Growth rate of R. pickettii QL-A6, R. pickettii QL-140 and the invader R. solanacearum did not increase linearly with increasing N : P ratio. While R. pickettii QL-140 had the highest growth rate when the N : P ratio was 0.16 (F_5,12 = 9.87, p < 0.001, electronic supplementary material, figure S5), R. pickettii QL-A6 grew the fastest at 16 and 160 N : P ratios (F_5,30 = 24.87, p < 0.001 and F_5,12 = 26.77, p < 0.001, respectively in the electronic supplementary material, figure S5). No difference in the maximum growth rate of these species was found at 1.6 N : P ratio (F_5,30 = 8.23, p < 0.001, electronic supplementary material, figure S5).

However, clear positive correlations were found between the productivity (population density after 72 h of growth) of the invader, QL-A6 and QL-140 species and the increasing N : P ratio (figure 3b). While R. pickettii QL-140 had the highest productivity at the lowest N : P ratio (0.16), the R. pickettii QL-A6 became more competitive relative to the invader at 16 and 160 N : P ratio treatments. This suggests that differences in species' ability to grow under increasing N : P ratios were probably important in explaining diversity–productivity–invasion resistance relationships.

Figure 3.

(a) The effect of N : P ratio on the resident community productivity–invasion resistance relationship. Resident community productivity was defined as optical density (OD₆₀₀) after 72 h incubation and shows the mean of all resident communities across all richness levels. The relative density of invader was defined as log-10 transformed relative mCherry fluorescence unit (RFU, mCherry/OD₆₀₀) after 72 h incubation. (b) The productivity of resident species and the invader at different N : P ratios measured in bacterial monocultures (OD₆₀₀ at 72 h of incubation).

(c). The effect of resource stoichiometry on the resident community productivity and invasions

We found that increasing N : P ratio had a hump-shaped (nonlinear) relationship with the resident community productivity (electronic supplementary material, figure S6, F_1,184 = 1.28, p = 0.26; the mean productivity of all resident communities in all richness levels), and only the resident community richness correlated positively with resident community productivity (model 5 in the electronic supplementary material, table S4). However, resident community productivity was positively affected by R. pickettii QL-A6 and QL-140 species and QL-A6 had a positive effect on community productivity with increasing N : P ratio (model 6 in the electronic supplementary material, table S4), which is in line with their ability to efficiently consume N and P (electronic supplementary material, figures S3 and S4) and to reach higher biomasses (productivity) with increasing N : P ratios in monocultures (figure 3b).

Resident community productivity had a clear negative effect on the relative density of the invader (figure 3a; model 4 in the electronic supplementary material, table S4), and crucially, the negative relationship between the resident community productivity and the density of the invader became stronger with increasing N : P ratio (figure 3a; model 4 in the electronic supplementary material, table S4). This can be explained by resident species identity effects, in particular, the ability of R. pickettii QL-A6 to increase its growth with increasing N : P ratio, which then turned the productivity–invasion resistance relationship more pronounced.

(d). Linking species identity and resource stoichiometry effects with productivity–invasion resistance relationship

To further study how invasions and community productivity were shaped by resource stoichiometry and species identity effects, we built an SEM describing direct and indirect relationships between these variables. The final SEM explained 76% of the variance of the relative density of the invader (figure 4). The species R. pickettii QL-A6 had a significant negative effect on the relative density of the invader, while the species R. pickettii QL-140 promoted community productivity and reduced the relative density of the invader. Similar to previous analyses, resource stoichiometry did not change the species identity effects in the SEM. However, resource stoichiometry had negative effects on resident community productivity and the relative density of invader, while the resident community productivity itself had a negative effect on invasions. Together these results suggest that species identity and resource stoichiometry had direct and indirect negative effects on invasions and that the indirect effects were mediated by resident community productivity.

Figure 4.

Structural equation model presenting direct and indirect effects of species identities and resource stoichiometry on resident community productivity and relative density of the invader. Resident community productivity was defined as optical density (OD₆₀₀) after 72 h incubation. Invasion success was defined as log-10 transformed relative mCherry fluorescence unit (RFU, mCherry/OD₆₀₀) after 72 h incubation. Continuous and dashed arrows indicate positive and negative effects, respectively, and the width of the arrows indicates relative effect sizes. Black circles indicate the proportion of the total variance explained and asterisks indicate significant effects (*p < 0.05; **p < 0.01; ***p < 0.001).

4. Discussion

Biodiversity is an important determinant of ecosystem functioning having significant effects on community resistance to biological invasions [5,42,43]. Here, we addressed how changes in environmental stoichiometry (N : P ratio) affect invasion resistance of model microbial communities. We found that changes in resource stoichiometry had clear effects on invasion outcomes via productivity-mediated species identity effects. First, increasing the N : P ratio lowered the intercept of diversity–invasion resistance relationship, which suggests that increasing the input of N reduced the likelihood of invasions regardless of the community diversity. Second, two resident species, R. pickettii QL-A6 and QL-140, played key roles in having negative effects on the invader and positive effects on resident community productivity. Crucially, increasing the N : P ratio turned the slope of the productivity–invasion resistance relationship much steeper because the species R. pickettii QL-A6 and R. pickettii QL-140 grew better and constrained invasions more efficiently when N became more abundant. Together these results suggest that resource stoichiometry can change the outcome of microbial invasions via productivity-mediated species identity effects.

In line with the previous studies, increasing resident community diversity decreased the likelihood of successful invasions [4,5,44]. While resource stoichiometry did not interact with resident species richness, it lowered the intercept of the diversity–invasions resistance relationship. This suggests that an increase in the relative concentration of N improved the resident community invasion resistance regardless of the species richness but that this effect was stronger in more diverse communities. One simple explanation for this is that increasing community diversity increased the likelihood that one or both of the species that were effective at constraining invasions (R. pickettii QL-A6 and QL-140) were included in communities. To study this in more detail, we concentrated on exploring the relative importance of resident species identities on invasions.

Two resident species, R. pickettii QL-A6 and QL-140, played key roles in having negative effects on the invader relative density and positive effects on the resident community productivity. Both of these species were effective at consuming N and P across all N : P ratios (electronic supplementary material, figures S3 and S4), and hence, their contribution to invasion resistance probably overshadowed the effects of the other resident community members. However, no difference was found in the consumption of N and P among the invader, R. pickettii QL-A6 and QL-140 (electronic supplementary material, figures S3 and S4), which suggests that these three species were equally efficient at sequestering N and P. However, either the R. pickettii QL-A6 or QL-140 was clearly faster at growing than the invader in three out of four N : P ratios used in our experiments (electronic supplementary material, figure S5), which could have helped them to outcompete the invader in these conditions. Moreover, while these species constrained invasions across all N : P ratios (figure 4), we found that the importance of species QL-A6 on community productivity increased along an increasing N : P ratio, while species QL-140 had the highest productivity at low N : P ratios (figure 3b). Together these results suggest that resource stoichiometry changed invasion outcomes via productivity-mediated species identity effects by favouring resident species that were efficient at growing when the N was abundant. This finding is in line with a previous study where these two species were observed to have highly negative effects on the same invader used in this study owing to high catabolic similarity [17] and supports the idea that individual contribution of resident community members on invasions can change according to resource availability [17]. In the future, it would be interesting to study if environmental stoichiometry can drive changes in the elemental stoichiometry of bacterial and other microbial cells. For example, it has been shown that the effects of resource stoichiometry can be species-specific [45] where environmental stoichiometry favours species with similar biomass composition [25,26]. In this case, the similarity in biomass composition between the resident species and the invader could be an important determinant for invasions.

Furthermore, we found that increasing the N : P ratio had a hump-shaped effect on community productivity that peaked at intermediate N : P ratios, which can optimize species coexistence or community productivity [19,22,46]. However, this relationship was not very strong and was only visible when all the communities with different richness levels were included in the analysis. Interestingly, resident community productivity correlated negatively with the relative density of the invader only within 16 and 160 N : P ratios. Mechanistically, this could be explained by the fact that the importance of R. pickettii QL-A6 on resident community productivity and invasions resistance increased along an increasing N : P ratio (figure 3b and model 2 in table 2). This suggests that increasing the input of N can increase the invasion resistance of communities via productivity, but that these effects might be driven by certain important ‘key stone’ species instead of changes in the total community productivity [9,17]. Several previous studies have suggested that resource stoichiometry of the environment is a good predictor of species growth capacity [45,47]. For example, N and P are important for species growth via effects on production and expression of proteins, enzymes and cell structures [27,48,49] and could often be limiting resources in the environment. Furthermore, it is possible that invasion resistance is mediated not only by N uptake but also by C metabolism, which is known to be interconnected with N regulation [50]. In support of this, a previous study has shown that the same resident species used in this study had higher growth rates, productivity and high resource niche overlap with the same invader used in this experiment when measured in various C media [4,17]. As a result, it is possible that competition for C and N affected the observed invasion outcomes also in this experiment.

Here, we link the high N : P ratio to improved community functioning in terms of increased invasion resistance. Our results suggest that resource stoichiometry can have positive effects on resident community productivity by favouring species that are very efficient at constraining invasions. Interestingly, resource stoichiometry did not change the shape of the diversity–invasion resistance relationship even though invasions were less successful in more diverse communities in general. By contrast, resource stoichiometry turned the slope of the productivity–invasion resistance more negative, because increase in N availability potentially intensified the competitive interactions between resident community members and the invader by favouring the growth of certain resident community members (R. pickettii QL-A6 and R. pickettii QL-140). This is in line with studies that showed that competition becomes stronger under a higher N : P ratio [28,51]. In the future, it will be important to better understand the effects of resource stoichiometry on invasions in more natural environments. For example, more information is needed as to how more complex microbial community, multi-trophic interactions with predators and parasites, root exudation and spatially uneven distribution of particulate organic matter shape the elemental stoichiometry and their effects on invasions in complex plant–soil ecosystems.

We conclude that resource stoichiometry is an important determinant of community invasion resistance. Human activities continue to have a huge effect on global elemental cycling [52], N leaching and eutrophication, which are causing growing problems and having devastating effects on the functioning of ecosystems [53,54]. In the case of eutrophication, our results suggest that resource stoichiometry could drive changes in microbial community composition potentially affecting the likelihood of biological invasions. In the agricultural context, resource stoichiometry could affect the severity of disease epidemics via effects on microbial competition. A better understanding of this process could potentially help to control plant pathogen invasions via modulation of soil nutrient availability and balance (N : P ratios) to maintain a relatively stable and invasion-resistant microbial community. In a broader perspective, understanding how changes in global element balances affect the interactions within and between communities is crucial for predicting ecosystem-level responses to environmental change.

Data accessibility

This article has no additional data.

Authors' contributions

A.J., Z.W. and T.Y. designed the experiment and T.Y. carried out the laboratory work with the help of G.H. and Q.Y. and analysed all the data. All authors wrote the manuscript and gave final approval for publication.

Competing interests

We declare we have no competing interests.

Funding

This research was financially supported by the National Natural Science Foundation of China (41471213 to Y.X.; 41671248 to Z.W.; 41807045 to T.Y.), National Key Basic Research Program of China (2015CB150503, Q.S.), the 948 project of the Ministry of Agriculture (2016-X45, YX), the Natural Science Foundation of Jiangsu Province (grant no. BK20170085, Z.W. and grant no. BK20180527, T.Y.), the 111 project (B12009, Q.S.), Young Elite Scientist Sponsorship Program by CAST (2015QNRC001 to Z.W.), the Qing Lan Project (funding to Y.X. and Z.W.), and the Chinese Scholarship Council (CSC) joint PhD scholarship. Ville-Petri Friman is supported by the Wellcome Trust [ref: 105624] through the Centre for Chronic Diseases and Disorders (C2D2) and Royal Society Research Grant (RSG\R1\180213) at the University of York. A.J. was supported by the Dutch Science foundation NWO (870.15.050) and the Koninklijke Nederlandse Akademie van Wetenschappen (530-5CDP18).