Divergence among species with “good competitor” and “good cultivator” strategies promotes asymmetric facilitation among co-invaders Authors: Chong-Wei Li, Hui-Xuan Liao, Ragan M. Callaway, Zhe-Yang Su, Ji-Kai Zou, Ao Liu, Yu-Ru Wu, Yu-Qi Fang, Shao-Lin Peng, Bao-Ming Chen Published: 2026 ABSTRACT Facilitative interactions among co-invaders may lead to invasional meltdown, accelerating non-native species accumulation and exacerbating ecological impacts over time. However, it remains unclear why certain non-native combinations promote facilitation while others do not, and may even constrain invasions. To address this question, we examine six invasive species in the Asteraceae family along two strategic dimensions: competitiveness and capacity to cultivate invader-promoting microbial communities. We then create experimental combinations to mix “good competitors” and “good cultivators” to varying degrees to form a “strategic divergence” gradient. We hypothesize greater strategic divergences generate more intense facilitations, whereas similar strategies generate inhibitions. Strategic divergence correlates with facilitation, but interactions are asymmetric: strong competitive suppressors of natives benefit from co-invasions with weaker competitors that cultivate favorable microbial environments but the performance of the latter are generally suppressed by the strong competitors. Metagenomic sequencing further indicates that good cultivators may promote facilitation by repelling pathogens (Ascomycota) and deterring microbes that might be exclusively beneficial for natives (Proteobacteria, Firmicutes, and Planctomycetota). Our results provide empirical evidence for the importance of strategic divergence among invasive species and offer a mechanistic basis for predicting which combinations of co-invading species might generate facilitation and which might result in inhibition. Introduction Global terrestrial ecosystems are facing exceptional pressure from biological invasions1 and co-invasion by multiple invasive species is becoming increasingly common2,3. Co-invading species may compete with each other4–7, but in some cases they facilitate each other, with either one or several performing better when co-occurring – a phenomenon that can lead to invasional meltdown8,9. Most evidence for meltdown comes from mutualistic relationships between plants and the animals that disperse or pollinate them8, but direct and indirect interactions among multiple co-invading plant species can also lead to meltdown6,10–13. However, why some combinations of invasive plant species generate facilitative interactions and other combinations do not, or even result in inhibition among co-invaders, is unknown. Thus, understanding the underlying mechanisms that drive facilitation among particular groups of invasive species, but not others, is essential for predicting the threats that co-invasions might pose to native ecosystems8,14,15. The leading hypothesis for invasions, co- or not, is escape from enemies16, and soil biota comprises one group of enemies that are embedded in soil communities which also include biota that can drive indirect facilitation among plants17. A growing body of evidence from biogeographically explicit plant-soil feedbacks (PSFs) indicates that escape from soil biota in native ranges can be crucial. For example, Maron et al.18 conducted PSF experiments with six native Eurasian species that are invasive in North America and reported strong negative PSFs in soil from the native range for five of the six species, indicating that soil biota from the native range suppresses these species. In contrast, no species had significant negative PSFs in soil from the non-native range. Similar biogeographically based evidence for escape from soil biota has been reported in other studies19–22. Recent studies have reported that PSFs can drive facilitation between two invaders11,23. This is because the abundance and diversity of pathogens on non-native plants can increase with the time since introduction24,25 and multi-species invasion can dilute the accumulation of soil pathogens26. This may release co-invaders from their own pathogenic enemies, consequently resulting in indirect facilitation among co-invaders. Escaping from harmful soil biota may open a door to meltdown, but beneficial soil biota may provide requisite indirect facilitation among co-invaders. Some invaders experience “enhanced mutualisms” with soil biota in non-native ranges27–31, whereas others disrupt mutualistic interactions between native plants and soil microbes32–34. Escape from inhibitory soil biota, linking up with more mutualistic soil biota, or diminishing the beneficial microbial effects on native species might generate net facilitative interactions among non-native invasive species, if in the process they cultivate soil biota that promotes other non-native plant species. Thus, in our quest to identify species or characteristics that might promote facilitation among co-invaders, we propose that species with strong microbial shaping abilities, or “good cultivators” should be key contributors to co-invader facilitation. “Good cultivators” may create a favorable microbial environment that indirectly promotes further invasion, but “good competitors” may suppress the growth of native species in a way that reduces the overall biotic resistance to further invasion35–37. Being a “good competitor” also appears to be a major constituent of successful invasions9,38–41. The stronger competitive ability of exotic invaders than native species may stem from rapid resource uptake and utilization15,42,43 or strong allelopathic potential44,45, which directly inhibits the growth of native species. This disproportionate competitive ability may be range-dependent, with experiments testing the competitive effects of invaders in their native versus non-native ranges in the field46 and against species from both ranges in common conditions47,48 showing much stronger competitive ability in the non-native scenarios. Many different combinations of functional strategies (e.g.,49,50,), other than good competitors and good cultivators, may result in facilitation among co-invaders, but clearly there is good reason to explore facilitation among combinations of species that vary in these strategies. First, these strategies contribute to single-species invasions51,52. Second, being a good competitor or a good microbial cultivator requires substantial resource allocations but for different purposes30,53,54. Strategic divergence, in general, could reduce niche overlap and again promote facilitation among co-invaders55,56 (Fig. 1). In other words, when multiple good competitors co-invade, high overlap in resource use may lead to strong competition among the invaders. When multiple good cultivators co-invade, they may play redundant roles in shaping the soil microbial community, and thus not offer additive benefits. Co-invasion involving the two different strategies may allow functional complementarity. Since net plant interactions are produced through a balancing act of relative intensities of competition and facilitation57–59, species should have a higher probability of facilitating each other if the contribution of competition to the net effect is weak. For co-invading species, good competitors may benefit from favorable microbial communities shaped by good cultivators and good cultivators may benefit from either moderate competition among co-invaders or reduced biotic resistance from native species interacting with good competitors.Fig. 1Hypothetical relationships between co-invasion outcomes and the strategic divergence among co-invaders.The black arrow represents the direction of increased strategic divergence between the ‘good competitor’ and ‘good cultivator’ strategies. Different invaders may diverge along these two dimensions or other strategic dimensions. Combinations of different strategists might reduce niche overlap and promote facilitation among co-invaders, while combinations of similar strategists might increase competition and promote inhibition. Hypothetical relationships between co-invasion outcomes and the strategic divergence among co-invaders. The black arrow represents the direction of increased strategic divergence between the ‘good competitor’ and ‘good cultivator’ strategies. Different invaders may diverge along these two dimensions or other strategic dimensions. Combinations of different strategists might reduce niche overlap and promote facilitation among co-invaders, while combinations of similar strategists might increase competition and promote inhibition. Here, we explore how strategic divergence among co-invaders might foster invasional facilitation. To do this, we conduct a controlled greenhouse experiment featuring six invasive species in the Asteraceae family (Supplementary Table 1) in which we determine how each varied across the “good competitor” and “good cultivator” strategic dimensions. We grow different combinations of these species, ensuring that each species was mixed with the other five species at least once, creating a gradient of strategic divergence across the combinations. We do this by establishing fourteen replicate assemblage mixtures – comprising a non-invaded native assemblage, six single-species-invaded assemblages and seven species combinations of co-invaded assemblages – and live vs. sterilized soil treatments (Fig. 2a). We ask the following questions: (i) Is facilitation more common than inhibition among co-invading species (Fig. 2b)? (ii) Is the magnitude of strategic divergence correlated with the severity of invasional facilitation among co-invaders (Fig. 2c, d)? And (iii) what microbial species are associated with invasional facilitation among co-invaders (Fig. 2e)?Fig. 2Experimental protocol and analysis.a The experiment involving two soil treatments (live vs. sterilized soil) and three invasion treatments (non-invaded native assemblage, single-species-invaded assemblages, and three-species-invaded assemblages). For invaded treatments, we adopted a replacement approach in which one native individual was replaced by one invasive individual. This approach is more conservative than additive approaches for measuring competition. b Strength of the “good competitor” strategy and the “good cultivator” strategy quantified under sterilized and live soil treatments, respectively, and strategic divergence quantified by the perimeter of the triangle defined by the constituent species located in the strategic space; c Co-invasion outcomes evaluated by non-additive effects (NAE); d Relationship between co-invasion outcomes and strategic divergence evaluated by general linear regression; e Determination of the role of key microbial groups on the basis of soil metagenomic sequencing. Here, component invaders refer to the individual species that comprise the co-invasion assemblages. All elements in this figure were created by the authors using Adobe Illustrator. Source data are provided as a Source Data file. Experimental protocol and analysis. a The experiment involving two soil treatments (live vs. sterilized soil) and three invasion treatments (non-invaded native assemblage, single-species-invaded assemblages, and three-species-invaded assemblages). For invaded treatments, we adopted a replacement approach in which one native individual was replaced by one invasive individual. This approach is more conservative than additive approaches for measuring competition. b Strength of the “good competitor” strategy and the “good cultivator” strategy quantified under sterilized and live soil treatments, respectively, and strategic divergence quantified by the perimeter of the triangle defined by the constituent species located in the strategic space; c Co-invasion outcomes evaluated by non-additive effects (NAE); d Relationship between co-invasion outcomes and strategic divergence evaluated by general linear regression; e Determination of the role of key microbial groups on the basis of soil metagenomic sequencing. Here, component invaders refer to the individual species that comprise the co-invasion assemblages. All elements in this figure were created by the authors using Adobe Illustrator. Source data are provided as a Source Data file. Results Divergence between “good competitor” and “good cultivator” strategies We first calculated competitor scores of our six invasive species by measuring the strengths of their competitive suppression of natives. This calculation utilized the competition index of single-species invasion in sterilized soil which quantified the competitive response of natives to invasions, the inverse of the competitive suppression of natives by each invasive species. The invader that suppressed native species the most was given a competitor score of 100, whereas the invader exhibiting the least suppression of natives received a score of 0. Across the five replicate measurements for each invader, we identified Tridax (87.1 ± 11.6), Bidens (80.9 ± 6.7), and Ageratum (57.1 ± 8.0) as good competitors, and Praxelis (29.6 ± 11.6), Sphagneticola (23.6 ± 7.1), and Chromolaena (17.0 ± 9.7) as weak competitors (Fig. 3a; Supplementary Table 2).Fig. 3Strategic ranks and the non-additive effect of co-invasion.a Competitor scores of invasive species, quantified by invader’s competitive suppression of native species; b Cultivator scores of invasive species, quantified by the compositional distance of the microbial community between invaded and non-invaded treatments. Comparisons of scores among different groups were based on one-way ANOVA, and significant differences in competitor and cultivator scores are marked by different letters based on the Tukey HSD multiple comparison test (P < 0.050). c Biomass changes of invasive species during co-invasions compared with those during single-species invasions; d Non-additive effects (NAE) of co-invasion on invader biomass. e NAE of co-invasion on native biomass; f NAE of co-invasion for the competition index of invasive species. In panel c, we also show the competitor (Comp) and cultivator (Cult) strategies exhibited by each invasive species (Sp). Significant changes in NAE or biomass are marked by asterisks based on two-sided t-tests (** P < 0.01, * P < 0.05, † P < 0.1). Means and standard errors are shown for all the graphs (n = 5 per species). Different species combinations were labeled using the first letters of the species names of the three focal constituent species: A for Ageratum conyzoides, B for Bidens alba, C for Chromolaena odorata, P for Praxelis clematidea, S for Sphagneticola trilobata, and T for Tridax procumbens. Full statistical details are available in Supplementary Information (Tables S2-5).Source data are provided as a Source Data file. Strategic ranks and the non-additive effect of co-invasion. a Competitor scores of invasive species, quantified by invader’s competitive suppression of native species; b Cultivator scores of invasive species, quantified by the compositional distance of the microbial community between invaded and non-invaded treatments. Comparisons of scores among different groups were based on one-way ANOVA, and significant differences in competitor and cultivator scores are marked by different letters based on the Tukey HSD multiple comparison test (P < 0.050). c Biomass changes of invasive species during co-invasions compared with those during single-species invasions; d Non-additive effects (NAE) of co-invasion on invader biomass. e NAE of co-invasion on native biomass; f NAE of co-invasion for the competition index of invasive species. In panel c, we also show the competitor (Comp) and cultivator (Cult) strategies exhibited by each invasive species (Sp). Significant changes in NAE or biomass are marked by asterisks based on two-sided t-tests (** P < 0.01, * P < 0.05, † P < 0.1). Means and standard errors are shown for all the graphs (n = 5 per species). Different species combinations were labeled using the first letters of the species names of the three focal constituent species: A for Ageratum conyzoides, B for Bidens alba, C for Chromolaena odorata, P for Praxelis clematidea, S for Sphagneticola trilobata, and T for Tridax procumbens. Full statistical details are available in Supplementary Information (Tables S2-5).Source data are provided as a Source Data file. We calculated the cultivator scores of the invasive species on the basis of the strength of microbial-shaping ability by quantifying the microbial compositional distance between the single-species-invaded and non-invaded assemblages under live soil conditions. The invasive species with the greatest compositional distance from non-invaded pots were given a cultivator score of 100, while the invader with the smallest distance from non-invaded pots were given a score of 0. Using the means of five measurements per species, we identified Bidens (79.9 ± 6.0), Tridax (63.0 ± 9.7) and Chromolaena (56.5 ± 11.6) as good cultivators, and Praxelis (27.2 ± 7.1), Ageratum (23.3 ± 7.7), and Sphagneticola (18.7 ± 5.0) as weak cultivators (Fig. 3b; and Supplementary Table 2). Subsequently, we quantified the competitor-cultivator strategic divergence for each three-species combination using the mean pairwise distance among the constituent species in competitor-cultivator strategic space (Supplementary Table 3). The rankings of strategic divergence across combinations were as follows: Tridax-Chromolaena-Sphagneticola (61.8), Bidens-Ageratum-Sphagneticola (59.5), Bidens-Praxelis-Chromolaena (57.6), Bidens-Praxelis-Sphagneticola (55.8), Tridax-Ageratum-Praxelis (48.2), Bidens-Tridax-Ageratum (43.0), and Ageratum-Chromolaena-Sphagneticola (41.2) (Supplementary Table 3). Invasional facilitation among co-invaders and strategic divergence No three-species combinations exhibited complete facilitative interactions among co-invaders in which all species benefited. Instead, some species benefited from co-invasions, whereas the performance of others diminished (Fig. 3c; and Supplementary Table 4). Specifically, Bidens and Tridax, which excelled at both the competitor and the cultivator strategies, consistently grew larger in co-invasions than in single-species invasions, except when they were grown together (see the Bidens-Tridax-Ageratum combinations in Fig. 3c). In contrast, Chromolaena, a weak competitor but good cultivator, consistently grew smaller in co-invasions than in single-species invasions. These results indicate common asymmetric facilitation among co-invaders, where good competitors benefit from co-invasions by compromising the performance of other co-invaders, particularly those with strong microbiota-cultivating abilities but weak competitive abilities. Notably, no invasive species benefitted from co-invasions when species with similar strategies were mixed. For example, in the combination of Bidens, Tridax, and Ageratum, which had the second lowest divergence score with all the component invaders being superior competitors, no invader significantly increased in biomass during co-invasions. Similarly, for the combination of Ageratum, Chromolaena, and Sphagneticola, which had the lowest divergence score, no invader significantly benefited from co-invasions. The strength of co-invader facilitation was further evaluated by calculating the non-additive effect (NAE) of co-invasion, which was quantified as the difference between the observed and expected species performance—the latter being approximated by the performance of invasive and native species in single-species invasions. Five of the seven co-invasion combinations exhibited significant or marginally significant positive NAE on the biomass of co-invaders (Fig. 3d; and Supplementary Table 5). Four of the seven co-invasion combinations exhibited significant or marginally significant negative NAE on the biomass of native species (Fig. 3e), whereas three combinations exhibited significant or marginally significant negative NAE on the competitive index (Fig. 3f). Thus, facilitation was common but complex (Fig. 3d–f). For example, the Bidens-Tridax-Ageratum combination exhibited facilitation in terms of invader biomass, but inhibition among co-invaders in terms of native biomass and the competition index. The Bidens-Ageratum-Sphagneticola combination exhibited facilitation measured as invader biomass, but only a weak directional trend towards inhibition in terms of competition index (NAEcompetition index = 0.08 ± 0.07, P = 0.302; Supplementary Table 5). The only combination that exhibited weak directional trends of inhibition for all metrics was the Ageratum-Chromolaena-Sphagneticola combination (NAEinvader biomass = -0.11 ± 0.25, P = 0.687; NAEnative biomass = 0.15 ± 0.13, P = 0.292; NAEcompetition index = 0.13 ± 0.08, P = 0.178; Supplementary Table 5). The level of competitor-cultivator strategic divergence was positively correlated with the NAE of co-invasion on invader growth (R2 = 0.58, P = 0.048, Fig. 4a) and it exhibited a trend of a negative correlation with the NAE of co-invasion on native species growth (R2 = 0.52, P = 0.067, Fig. 4b). However, no significant correlation was observed between competitor-cultivator strategic divergence and the NAE of the competitive index (R2 = 0.37, P = 0.145, Fig. 4c).Fig. 4Linear regressions of co-invasion outcomes with competitor-cultivator strategic divergence.a Competitor-cultivator divergence and the non-additive effect (NAE) on invader biomass; b Competitor-cultivator divergence and the NAE on native biomass; c Competitor-cultivator divergence and the NAE on the competition index. Significant or marginally significant correlations are shown as solid lines (P < 0.10), and non-significant correlations are shown as dashed lines. Each data point represents the mean of replicate measurements (n = 5) and the shaded areas in gray are 95% confidence intervals. Different species combinations were labeled using the first letters of the species names of the three focal constituent species: A for Ageratum conyzoides, B for Bidens alba, C for Chromolaena odorata, P for Praxelis clematidea, S for Sphagneticola trilobata, and T for Tridax procumbens. Source data are provided as a Source Data file. Linear regressions of co-invasion outcomes with competitor-cultivator strategic divergence. a Competitor-cultivator divergence and the non-additive effect (NAE) on invader biomass; b Competitor-cultivator divergence and the NAE on native biomass; c Competitor-cultivator divergence and the NAE on the competition index. Significant or marginally significant correlations are shown as solid lines (P < 0.10), and non-significant correlations are shown as dashed lines. Each data point represents the mean of replicate measurements (n = 5) and the shaded areas in gray are 95% confidence intervals. Different species combinations were labeled using the first letters of the species names of the three focal constituent species: A for Ageratum conyzoides, B for Bidens alba, C for Chromolaena odorata, P for Praxelis clematidea, S for Sphagneticola trilobata, and T for Tridax procumbens. Source data are provided as a Source Data file. Key microbial genera involved in facilitations among co-invaders Facilitation among co-invaders might be driven by an increase in microbes that benefit invaders or harm natives, and inhibition might be driven by microbes that harm invaders or benefit natives. To differentiate the roles of microbial species, we classified genera as invasion-facilitating if their abundances were positively correlated with invader biomass or negatively correlated with the competition index during single-species invasions. Conversely, genera were classified as invasion-inhibiting if the opposite patterns were observed. For each co-invasion combination, we identified candidate genera that associated with at least one constituent invader. Therefore, simultaneous increases in invasion-facilitating candidates or simultaneous decreases in invasion-inhibiting candidates during both single-species invasions and co-invasions indicate an improved microbial environment induced by constituent invaders. A larger fraction of microbial genera exhibiting such changes suggests a greater potential for these constituent invaders to create facilitation-promoting microbial environments. A significant proportion of invasion-inhibiting genera decreased simultaneously during both single-species invasions and co-invasions. Across different co-invasion combinations, these proportions ranged from 33.8% to 59.7% (Fig. 5b; and Table 1; Supplementary Fig. 1). Similarly, a considerable proportion of invasion-facilitating genera increased simultaneously during both single-species invasions and co-invasions, with proportions ranging from 14.9% to 37.1% (Fig. 5a; and Table 1; Supplementary Fig. 1).Fig. 5Fractions of invasion-related microbial genera and co-invasion outcomes.a Fraction of increased invasion-facilitating microbes that were induced by constituent invaders within each co-invasion combination, with the focal fraction highlighted in yellow in the Venn diagram subplot; b Fraction of decreased invasion-inhibiting microbes that were likely induced by constituent invaders within each co-invasion combination, also basing on the Venn-diagram approach; c Correlations between the non-additive effects (NAE) on invader biomass and the fraction of increased invasion-facilitating microbes; d Correlations between the NAE on invader biomass and the fraction of decreased invasion-inhibiting microbes; e Correlations between the NAE on native biomass and the fraction of increased invasion-facilitating microbes; f Correlations between the NAE on native biomass and the fraction of decreased invasion-inhibiting microbes; g Correlations between the NAE on the competition index and the fraction of increased invasion-facilitating microbes; h Correlations between the NAE on the competition index and the fraction of decreased invasion-inhibiting microbes. In panels a and b, each overlapping oval in yellow represents the fraction of invasion-influencing microbes that potentially increased or decreased due to the involvement of a consitutent co-invader. The sum of these fractions reflects the overall potential of the co-invaders to create the favorable microbial environments for facilitation among co-invaders. Linear regression was adopted to determine the correlation between the changed fraction of key microbes and invasion outcomes, which was conducted for the data of all seven co-invasion combinations (black line), as well as the data of six combinations excluding the Bidens-Tridax-Ageratum (BTA) combination (blue line). Significant or marginally significant correlations are shown as solid lines (P < 0.10), and non-significant correlations are shown as dashed lines. Each data point represents the mean of replicate measurements (n = 5) and the shaded areas in gray or blue are 95% confidence intervals. Different species combinations were labeled using the first letters of the species names of the three focal constituent species: A for Ageratum conyzoides, B for Bidens alba, C for Chromolaena odorata, P for Praxelis clematidea, S for Sphagneticola trilobata, and T for Tridax procumbens. Source data are provided as a Source Data file.Table 1Proportions and counts of invasion-facilitating and -inhibiting microbes potentially induced by constituent invaders within each co-invasion combinationCombinationsIncreased invasion-facilitating fractionDecreased invasion-inhibiting fractionTCS25.6%(63/213)54.7%(140/243)TAP15.0%(31/141)46.8%(90/161)BPC37.1%(72/135)59.7%(310/437)BPS16.6%(60/244)41.3%(204/359)BAS14.9%(63/356)33.8%(170/418)ACS24.3%(59/242)39.5%(199/251)BAT20.5%(58/191)43.2%(178/285)The numbers before brackets indicate the relative proportions of genera that simultaneously changed during both co-invasions and single-species invasions averaged across replicate measurements (n = 5). The numbers in brackets before the “/” indicate the counts of these genera. The numbers in brackets after the “/” represent the sums of all the genera that changed in association with the constituent invaders, either during single-species invasions or co-invasions. Different species combinations were labeled using the first letters of the species names of the three focal constituent species: A for Ageratum conyzoides, B for Bidens alba, C for Chromolaena odorata, P for Praxelis clematidea, S for Sphagneticola trilobata, and T for Tridax procumbens. Fractions of invasion-related microbial genera and co-invasion outcomes. a Fraction of increased invasion-facilitating microbes that were induced by constituent invaders within each co-invasion combination, with the focal fraction highlighted in yellow in the Venn diagram subplot; b Fraction of decreased invasion-inhibiting microbes that were likely induced by constituent invaders within each co-invasion combination, also basing on the Venn-diagram approach; c Correlations between the non-additive effects (NAE) on invader biomass and the fraction of increased invasion-facilitating microbes; d Correlations between the NAE on invader biomass and the fraction of decreased invasion-inhibiting microbes; e Correlations between the NAE on native biomass and the fraction of increased invasion-facilitating microbes; f Correlations between the NAE on native biomass and the fraction of decreased invasion-inhibiting microbes; g Correlations between the NAE on the competition index and the fraction of increased invasion-facilitating microbes; h Correlations between the NAE on the competition index and the fraction of decreased invasion-inhibiting microbes. In panels a and b, each overlapping oval in yellow represents the fraction of invasion-influencing microbes that potentially increased or decreased due to the involvement of a consitutent co-invader. The sum of these fractions reflects the overall potential of the co-invaders to create the favorable microbial environments for facilitation among co-invaders. Linear regression was adopted to determine the correlation between the changed fraction of key microbes and invasion outcomes, which was conducted for the data of all seven co-invasion combinations (black line), as well as the data of six combinations excluding the Bidens-Tridax-Ageratum (BTA) combination (blue line). Significant or marginally significant correlations are shown as solid lines (P < 0.10), and non-significant correlations are shown as dashed lines. Each data point represents the mean of replicate measurements (n = 5) and the shaded areas in gray or blue are 95% confidence intervals. Different species combinations were labeled using the first letters of the species names of the three focal constituent species: A for Ageratum conyzoides, B for Bidens alba, C for Chromolaena odorata, P for Praxelis clematidea, S for Sphagneticola trilobata, and T for Tridax procumbens. Source data are provided as a Source Data file. Table 1Proportions and counts of invasion-facilitating and -inhibiting microbes potentially induced by constituent invaders within each co-invasion combinationCombinationsIncreased invasion-facilitating fractionDecreased invasion-inhibiting fractionTCS25.6%(63/213)54.7%(140/243)TAP15.0%(31/141)46.8%(90/161)BPC37.1%(72/135)59.7%(310/437)BPS16.6%(60/244)41.3%(204/359)BAS14.9%(63/356)33.8%(170/418)ACS24.3%(59/242)39.5%(199/251)BAT20.5%(58/191)43.2%(178/285)The numbers before brackets indicate the relative proportions of genera that simultaneously changed during both co-invasions and single-species invasions averaged across replicate measurements (n = 5). The numbers in brackets before the “/” indicate the counts of these genera. The numbers in brackets after the “/” represent the sums of all the genera that changed in association with the constituent invaders, either during single-species invasions or co-invasions. Different species combinations were labeled using the first letters of the species names of the three focal constituent species: A for Ageratum conyzoides, B for Bidens alba, C for Chromolaena odorata, P for Praxelis clematidea, S for Sphagneticola trilobata, and T for Tridax procumbens. Proportions and counts of invasion-facilitating and -inhibiting microbes potentially induced by constituent invaders within each co-invasion combination The numbers before brackets indicate the relative proportions of genera that simultaneously changed during both co-invasions and single-species invasions averaged across replicate measurements (n = 5). The numbers in brackets before the “/” indicate the counts of these genera. The numbers in brackets after the “/” represent the sums of all the genera that changed in association with the constituent invaders, either during single-species invasions or co-invasions. Different species combinations were labeled using the first letters of the species names of the three focal constituent species: A for Ageratum conyzoides, B for Bidens alba, C for Chromolaena odorata, P for Praxelis clematidea, S for Sphagneticola trilobata, and T for Tridax procumbens. As predicted, a greater fraction of invasion-inhibiting microbes that decreased simultaneously during both single-species invasions and co-invasions were associated with more intense invasional facilitation among co-invaders, as evidenced by decreased NAE on native biomass and the competition index (Fig. 5e–h). These trends became more prominent after excluding the Bidens-Tridax-Ageratum combination, in which all invaders were good competitors, based on regression analysis (see solid blue lines in Fig. 5f, h). This suggests that the reduction in invasion-inhibiting microbes may be crucial for invasional facilitation mediated by good cultivators. Among all the phyla, Ascomycota, Proteobacteria, Firmicutes, Planctomycetota, and Uroviricota emerged as key phyla containing many invasion-inhibiting genera whose abundance decreased during both single-species invasions and co-invasions (Fig. 6; and Supplementary Data 1). This suggests that Ascomycota, Proteobacteria, Firmicutes, Planctomycetota, and Uroviricota may have dictated the co-invasion outcomes in our study system.Fig. 6Counts of microbes potentially induced by constituent invaders within each co-invasion combination.a Fraction of invasion-facilitating microbes; b Fraction of invasion-inhibiting microbes. Only the ten phyla with the greatest numbers of changed genera are shown for each fraction. Different species combinations were labeled using the first letters of the species names of the three focal constituent species: A for Ageratum conyzoides, B for Bidens alba, C for Chromolaena odorata, P for Praxelis clematidea, S for Sphagneticola trilobata, and T for Tridax procumbens. Source data are provided as a Source Data file. Counts of microbes potentially induced by constituent invaders within each co-invasion combination. a Fraction of invasion-facilitating microbes; b Fraction of invasion-inhibiting microbes. Only the ten phyla with the greatest numbers of changed genera are shown for each fraction. Different species combinations were labeled using the first letters of the species names of the three focal constituent species: A for Ageratum conyzoides, B for Bidens alba, C for Chromolaena odorata, P for Praxelis clematidea, S for Sphagneticola trilobata, and T for Tridax procumbens. Source data are provided as a Source Data file. Discussion “Invasional meltdown” occurs when one invasive species facilitates the likelihood of survival or increasing ecological impact of another8,60, a pattern consistent with our results. However, our short-term experiments do not demonstrate the positive feedback loop necessary to sustain a continuous meltdown. Instead, our results indicate that combinations of invasive species that diverge more along just two strategic dimensions, strong suppressors of natives vs. effective cultivators of microbial communities, are more likely to produce facilitation among co-invaders than species that diverge less (Fig. 4). In contrast, combinations of invaders with similar strategies exhibited no facilitation (BTA and ACS in Fig. 3c), underscoring the importance of strategic divergence in promoting co-invasions and potential invasional meltdown over a longer time scale. While other strategic dimensions, such as resource utilization efficiency and allelopathy, may also underlie direct facilitation among invaders, their influence remains to be explored. Importance of strategic divergence in co-invader facilitation Consistent with previous studies that focused on interactions among native plants61–64, our results underscore the crucial role of functional or strategic divergence in promoting interspecific facilitation across invasive species (Fig. 3c and Fig. 4a, b). Increased strategic divergence allows functional complementarity and reduces niche overlap among co-existing species65,66. This may increase community-level resource uptake and space occupation, thereby increasing community productivity and impeding establishment of newcomers67,68. For example, divergence in leaf economic traits, reflecting strategic divergence in resource uptake and utilization, plays a key role in facilitating interactions among natives66,69,70. These results highlight the significant roles of complementarity and selection effects as fundamental mechanisms for promoting facilitative interactions among native species via strategic divergence. Like native species, strategic divergence may also allow stronger facilitative interactions among invaders. However, to our knowledge, no prior study has considered how strategic divergence may aid invasion or amplify the impact of invasion. Role of “mutualism disruption” in co-invader facilitation Through metagenomic sequencing, we identified key microbial genera associated with co-invader facilitation. Unlike previous emphasis on the role of fungal pathogens11,13, our results suggest that “disrupting mutualisms” through diminishing microbes more beneficial for natives than invasive species might be crucial. Invasive plants have been reported to disrupt mutualistic interactions between native plants and soil microbes32–34,71, which might promote invasional meltdowns. For example, Callaway et al.71 and Roche et al.72 reported that invasive Alliaria petiolata produced allelochemicals that disrupted the mutualistic associations between native herbs and mycorrhizal fungi, which put native mycorrhizal herbs at a competitive disadvantage. Similarly, we found that invaders reduced the abundances of Proteobacteria, Firmicutes, Planctomycetota, and Ascomycota, which all have positive effects on native resistance to invasions (Supplementary Data 1). Proteobacteria and Firmicutes suppress diseases73–75, and Planctomycetota and Ascomycota accelerate carbon and nitrogen cycling76–79. The reduced abundance of these microbes may have led to greater disease incidence and nutrient limitation for native plants. Role of “enemy release” in co-invader facilitation “Enemy release” may also contribute to co-invader facilitation. Some invasive plants may gain a competitive advantage over local native species by escaping specialized pathogens16,19,20,46. Reduced pathogen densities80 and altered pathogen community composition81 have been reported to result in weaker pathogenic effects of soil biota on invaders in the non-native vs. native ranges. Given that Ascomycota contains many endophytic plant pathogens82, the significant reductions in Ascomycota pathogens during both single-species invasions and co-invasions may also contribute to the observed facilitative interactions among co-invaders. Prevalance of asymmetric facilitation Interestingly, we found that strong competitive suppressors of natives always benefited from co-invading with weaker competitors which created favorable microbial environments for invasion (Fig. 3c; and Fig. 5). Even though some invaders appeared to excel at both the competitor and cultivator strategies, these two strategies may often require a tradeoff in resource allocation (per Grime49). Both direct competition and microbiota cultivation demand significant energy investment30,53,54, but good competitors may take advantage of improved microbial environments created by co-invaders without having to reallocate resources to cultivating microbial communities. Alternatively, good cultivators may gain nothing from neighbors that allocate to competitive ability because they are subject to these neighbor’s competitive effects. Different scenarios of resource reallocation, such as reallocation between growth and herbivory defense or symbiosis, may contribute to enhanced plant performance83,84. However, to our knowledge no previous evidence has indicated that species with divergent allocational strategies can promote meltdown. Invaders were not restricted to excelling in only one strategy (Fig. 3), but we did find that co-invasion outcomes are promoted by a balance between indirect facilitative and competitive processes. That is, facilitation may occur only when the benefit of microbiota cultivation is stronger than competition among co-invaders. Caveats of the current study First, our calculation of the strength of invasional facilitation among co-invaders is based on NAE, which cannot differentiate mutual from asymmetric facilitation. Because the NAE only measures the overall deviation from single-species invasions, a significant NAE does not necessarily mean that all invaders benefit from co-invasions (see Fig. 3d–f vs. Fig. 3c). Such asymmetry may not lead to long-term coexistence among the invaders, but stronger negative impacts on native assemblages could still result in the competitive displacement of native species, which aligns with the broader concept of invasional meltdown. Second, we could not distinguish between the effects of increasing exotic diversity and decreasing intraspecific competition. Our experimental design does not allow us to test whether increased exotic diversity reduces species-specific pathogen accumulation and resource competition, which may act in conjunction to promote facilitation among co-invaders. Third, although the six invaders overlap biogeographically in their native ranges, they share no such history with the three native species, ruling out invader-native coevolution and ensuring synergistic effects do not confound our results. Lastly, we focused on the three most abundant native species, representing a subset of the local community. While this standard approach captures dominant vegetation, rarer natives might respond differently; future field experiments with greater native diversity would test the generality of our findings. Synthesis and conclusions In summary, we used three-invader study systems to study how divergence in strategies along axes of competition and cultivation of soil microbes affect invasional facilitation among co-invaders. We found that asymmetrical facilitation among co-invaders is prevalent and correlated with strategic divergence. Only “good competitors” reaped substantial benefits from co-invasion, when co-invading with invasive species that excelled at cultivating invader-friendly microbial communities but had limited competitive ability. These cultivators repel fungal pathogens and deter microbes that favored native over invasive species. We propose that this “Strategic Divergence Hypothesis” for facilitation among co-invaders might ultimately allow generalization of how co-invaders increase or decrease their success. However, our study was limited to invasive species in the Asteraceae and three native species—Emilia sonchifolia (Asteraceae), Solanum nigrum (Solanaceae), and Plantago asiatica (Plantaginaceae). These species are ecologically representative of the secondary shrub-grasslands in southern China, but they restrict broader generalization of our findings. Future research should consider the following directions to build upon our work: (1) Expanding taxonomic diversity of invasive species. The inclusion of more distantly related invasive species might generate stronger facilitative interactions among co-invaders, given that phylogenetic distance promotes facilitation85. (2) Increasing co-invasion combinations and strategic dimensions: The limited number of co-invasion combinations tested may explain the few significant inhibitions we detected. Expanding the number of combinations and exploring alternative axes of strategic divergence, such as resource-use efficiency or phenological traits, could provide deeper insights into the conditions that drive inhibitive interactions among co-invaders. This would complement our current focus on facilitative dynamics. (3) Incorporating diverse native species: Future studies should include a wider range of native species from more varied taxonomic groups and functional types, including woody plants or perennial herbs. This would better capture the complexity of invasive-native interactions across different ecological contexts and provide a more comprehensive understanding of the dynamics at play. By addressing these limitations, future research can develop a more robust understanding of the mechanisms underlying facilitation and inhibition among co-invaders, which play crucial roles in driving invasional meltdowns. Methods Study site and species A glasshouse experiment was conducted at the Ecological Research Station in Haizhu Wetland National Park, Guangdong, China (23°04’38”N, 113°18’9”E). Six invasive plant species native to the Americas (Bidens alba, Tridax procumbens, Ageratum conyzoides, Praxelis clematidea, Chromolaena odorata and Sphagneticola trilobata) and three native species commonly found in southern China (Emilia sonchifolia, Solanum nigrum and Plantago asiatica) were selected for this study (geographic distributions verified through Plants of the World Online, URL: https://powo.science.kew.org/) (Supplementary Table 1). Experimental design Owing to inherently poor seed germination, S. trilobata was propagated by stem cuttings. The seeds of the other eight study species were collected in March 2023, air-dried, and stored at room temperature. In June 2023, the seeds of these species were sown in trays filled with a 1:1 (v:v) mixture of sterilized perlite and nutrient-enriched soil. To ensure that the seedling sizes of all the plants were similar during transplanting, we adjusted the starting time of germination for each species to account for species-specific seed germination timing. In July 2023, seedlings of the same size were selected for transplanting. The substrate used for planting was a mixture of field soil, sterilized river sand, and perlite at a 3:2:1 (v:v:v) ratio. Because soil from the invaders’ non-native range was essential for capturing the shared microbial benefits among co-invaders that arise following enemy release and enhanced mutualism, the field soil used in this experiment was collected from Taizikeng Forest Park, Zengcheng District, Guangzhou city, Guangdong Province (23°18'25“N, 113°45'12“E). This site is characterized by a typical southern subtropical secondary shrub-grassland ecosystem, which is within the natural distribution range of the three native plant species we tested: Emilia sonchifolia, Solanum nigrum, and Plantago asiatica. These three native species were selected from the 18 herbaceous plant species recorded within a 50-meter radius of the soil sampling location; notably, these three species were dominant and formed a stable, moderately diverse community with other co-occurring native herbs. The nutrient content of the field soil was as follows: total nitrogen, 0.20 g kg⁻¹; total phosphorus, 0.14 g kg⁻¹; ammonium, 3.94 mg kg⁻¹; and nitrate, 4.14 mg kg⁻¹. Notably, that the microbial species pool may vary substantially across sites where the three native species co-occur. The limited soil origin in our study does not account for variation in microbial taxa. We established two soil microbial treatments (live and sterilized soil) and 14 plant assemblage treatments, including one native (non-invaded) plant assemblage, six single-species invasion treatments, and seven three-species co-invasion treatments (Fig. 2, and Supplementary Table 1). For the live-soil treatment, field soil was mixed with sand and perlite. For the sterilized treatment, the soil mixture was sterilized using 60Co irradiation at 50 kGy. The growth medium was then placed into 5-L pots (22 cm diameter, 15 cm height). For the non-invaded treatment, two individuals of each of the three native species were planted along the diagonals of a regular hexagon. For the invasion treatment, one individual of each native species was replaced by one invasive plant individual. Single-species invasion involved three individuals of the same invasive species, and co-invasion involved three individuals of three different invasive species. A set of seven co-invasion combinations was established to ensure that each invasive species co-invades with the other five invasive species in at least one replicate. We preferred this three-invader system over two-invader system for two reasons: (1) Three-species combinations allowed us to cover a broader and more continuous gradient of strategic divergence than two-species combinations. This approach provides a more comprehensive understanding of the interactions and effects we are studying. (2) The use of three-species combinations may reduce the potential for species-specific biases that might be stronger in two-species systems. Our experimental design ensured that each of the six invader co-occurs at least once with the other invaders, allowing for cross-validation and a more robust assessment of the effects. The only other way to have done this with six species would have been a very large experimental design. Details of species combinations are in Supplementary Table 1. The planting positions of the three invasive plants in each co-invasion treatment were randomly chosen. In total, the experiment included 2 soil treatments (live vs. sterilized) × 14 plant assemblage treatments × 5 replicates, resulting in a total of 140 pots with 840 individual plants. Sterile ultrapure water ( > 18.0 MΩ·cm) was used for daily irrigation throughout the experiment, approximately to the carrying capacity. A total of 50 ml of Hoagland’s solution was added per pot each week. Pot positions and orientations were changed weekly to homogenize environmental conditions. The average temperature during the experiment was 29.6 °C. In September 2023, two months after transplantation, the aboveground biomass of all plants was harvested, oven-dried at 70 °C for 48 h, and weighed. Soil sampling, DNA extraction, library construction, and metagenomic sequencing ~10 g of soil was collected from each pot, placed in 15 ml centrifuge tubes, and stored at −80 °C for soil microbial metagenomic sequencing. In total, we had 14 plant treatments × 5 replicates, resulting in 70 soil samples. Total genomic DNA was extracted from soil samples using the E.Z.N.A.® Soil DNA Kit (Omega Biotek, Norcross, GA, USA) following the manufacturer’s protocol. DNA concentration and purity were assessed with a TBS-380 fluorometer and a NanoDrop 2000 spectrophotometer, respectively. The integrity of the extracted DNA was verified via 1% agarose gel electrophoresis. For paired-end library construction, DNA was sheared to an average fragment size of ~400 bp using a Covaris M220 (Gene Company Limited, China) and processed with the NEXTFLEX Rapid DNA-Seq Kit (Bioo Scientific, Austin, TX, USA). Adapters containing sequencing primer hybridization sites were ligated to the blunt-ended fragments. Paired-end sequencing was performed on an Illumina NovaSeq 6000 (Illumina Inc., San Diego, CA, USA) at Majorbio Bio-Pharm Technology Co., Ltd. (Shanghai, China) using NovaSeq Reagent Kits, following the manufacturer’s instructions (www.illumina.com). The sequencing data from this project have been deposited in the NCBI database (Accession Number: PRJNA1187364 . Sequence quality control and microbial genome assembly The data were analysed on the Majorbio Cloud (www.majorbio.com) online platform. Briefly, paired-end Illumina reads were processed by fastp86 (version 0.20.0), where adaptors were trimmed, and low-quality reads (length <50 bp, quality score <20, or containing ambiguous bases) were removed (https://github.com/OpenGene/fastp, version 0.20.0). Metagenomic data were assembled via MEGAHIT87 (https://github.com/voutcn/megahit, version 1.1.2), which utilizes succinct de Bruijn graphs. Contigs ≥ 300 bp in length were selected for final assembly and used for subsequent gene prediction and annotation. Microbial taxonomy and species annotation A nonredundant gene catalog was generated using CD-HIT88 (http://www.bioinformatics.org/cd-hit/, version 4.6.1) with 90% sequence identity and 90% coverage. High-quality reads were aligned to the nonredundant gene catalogs to calculate gene abundance with 95% identity via SOAPaligner89 (http://soap.genomics.org.cn/, version 2.21). For taxonomic annotations, representative sequences from the nonredundant gene catalog were aligned to the NR database (NCBI nonredundant protein database, version 202209) using Diamond (version 0.8.35) with an e-value cutoff of 10-590 (http://www.diamondsearch.org/index.php, version 0.8.35). With respect to soil metagenomic sequencing, the accuracy of genus-level annotation is substantially greater than that of species-level91,92, thus we annotated soil microbes to the genus level. The dataset included annotated bacteria, fungi, archaea, viruses, and nematodes present in the soil, excluding other animals and plants. For different purposes, we constructed two distinct datasets. For community composition analysis, we constructed and utilized the relative number of reads to assess overall microbial community structure. To identify key microbes, we employed the Transcripts Per Million algorithm, which accounts for gene length and sequencing depth, providing a more accurate measure of microbial abundance. Statistical analyses All statistical analyses and plotting were conducted in R version 4.3.193. The final figures and legend modifications were performed with Adobe Illustrator 2024 (64-bit, https://www.adobe.com/). Competitive suppression of native species by invaders For each invasive species or combination of invasive species, the strength of competitive suppression of native species was calculated using the competition index (CI)94. For the i-th replicate pot, CI was calculated as follows:1\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{CI}}_{i}=\,\frac{2\times {{Biomass}}_{{invaded},i}}{{{Biomass}}_{{invaded},i}+\overline{{{Biomass}}_{{non}-{invaded}}}}$$\end{document}CIi=2×Biomassinvaded,iBiomassinvaded,i+Biomassnon−invaded¯where \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{Biomass}}_{{invaded},i}$$\end{document}Biomassinvaded,i is the total biomass of the native species in the i-th pot invaded by the focal invasive species, and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\overline{{{Biomass}}_{{non}-{invaded}}}$$\end{document}Biomassnon−invaded¯ is the mean total biomass of the native species of all non-invaded pots. Since our study focused on evaluating the inherent competitive ability of invasive species, the data for these calculations were derived from pots in the sterilized treatment. A competition index (CI) value closer to 0 indicates stronger competitive suppression of invaders on native species, whereas a CI value closer to 1 reflects weaker suppression. Co-invasion outcome In accordance with the invasional meltdown framework, the accumulation of invader abundance was assessed by changes in invader biomass while the accumulation of invader impact was assessed by changes in native biomass and competition index. Invader diversity was not included as a measure because our experimental design did not support its inclusion. Furthermore, we detected a significant positive correlation between the reduction in native biomass and the increase in invader biomass (Pearson’s R = 0.74, P < 0.001; Supplementary Fig. 2). This pattern indicates that greater negative impacts on native communities correspond to higher invader performance—directly confirming that our results align with the definition of invasional meltdown. To detect invasional facilitation and inhibition, we quantified the non-additive effects (NAEs) of co-invasion by comparing the observed performance of multiple invasive plants with their expected performance under co-invasion, which was derived from their measured performance under single-species invasion. The NAE on invader biomass was calculated via the following formula:2\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{NAE}}_{{invader}\,{biomass}}=\frac{{B}_{{observed}}-{B}_{{expected}}}{{B}_{{expected}}}$$\end{document}NAEinvaderbiomass=Bobserved−BexpectedBexpected3\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${B}_{{expected}}=\left({B}_{{inv}1}+{B}_{{inv}2}+{B}_{{inv}3}\right)/3$$\end{document}Bexpected=Binv1+Binv2+Binv3/3 Here, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${B}_{{observed}}$$\end{document}Bobserved represents the total aboveground biomass of the three constituent invasive species under co-invasion, and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${B}_{{expected}}$$\end{document}Bexpected represents the expected total aboveground biomass of the three invasive species under co-invasion based on the biomass of each invasive species under single invasion. \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${B}_{{inv}1}$$\end{document}Binv1 represents the aboveground biomass of the first constituent invader in single invasion. \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${B}_{{inv}2}$$\end{document}Binv2 represents the aboveground biomass of the second constituent invader in single invasion. \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${B}_{{inv}3}$$\end{document}Binv3 represents the aboveground biomass of the third constituent invader in single invasion. Similarly, the NAE on the native biomass was determined using the formula above. The NAE was calculated as follows:4\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{NAE}}_{i{nvader}\,{competitiveness}}=({{CI}}_{{observed}}-{{CI}}_{{expected}})$$\end{document}NAEinvadercompetitiveness=(CIobserved−CIexpected)5\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{CI}}_{{expected}}=\left(C{I}_{{inv}1}+{{CI}}_{{inv}2}+{{CI}}_{{inv}3}\right)/3$$\end{document}CIexpected=CIinv1+CIinv2+CIinv3/3 Here, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{CI}}_{{observed}}$$\end{document}CIobserved represents the competition index (CI) of invaders during co-invasion, and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{CI}}_{{expected}}$$\end{document}CIexpected represents the expected CI of the three invasive species in co-invasion based on the index of each invasive species in single invasion. \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{CI}}_{{inv}1}$$\end{document}CIinv1 represents the CI of the first constituent invader in single invasion. \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{CI}}_{{inv}2}$$\end{document}CIinv2 represents the CI of the second constituent invader in single invasion. \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{CI}}_{{inv}3}$$\end{document}CIinv3 represents the CI of the third constituent invader in single invasion. For invader biomass, a more positive NAE value indicates a more intense facilitation among co-invaders, while a more negative NAE value indicates a more intense inhibition. Conversely, for invader competitiveness and native biomass, a more negative NAE value indicates more intense facilitation among co-invaders, while a more positive NAE value indicates more intense inhibition. A two-tailed t test was then conducted to compare these NAEs against zero, with P < 0.05 indicating significant effects. The strength of the ‘good competitor’ and ‘good cultivator’ strategies We quantified the strength of the ‘good competitor’ strategy using the competitor-score metric, which is based on the strength of competitive suppression of the focal invader on native species (i.e., the CI of the focal species). Because a lower CI value indicates stronger competitive ability, we quantified the competitor score by translating the CIs onto an axis with the maximum value set as zero and the minimum value as 100:6\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{Competitor}\,{score}}_{i}=\frac{{{m}{a}{x}(CI){-}{C}{I}}_{i}}{max\left({CI}\right)-min\left({CI}\right)}\times 100$$\end{document}Competitorscorei=max(CI)−CIimaxCI−minCI×100 For each species, the strength of the ‘good cultivator’ strategy was quantified using a cultivator-score metric, which was based on the microbial shaping ability (MSA) of each species. For the i-th replicate pot, the MSA of the focal invasive species can be quantified as follows:7\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{MSA}}_{i}=\overline{{{Distance}}_{{invaded}\,{vs}.\,{non}-{invaded}}}$$\end{document}MSAi=Distanceinvadedvs.non−invaded¯where \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\overline{{{Distance}}_{{invaded\; vs}.{non}-{invaded}}}$$\end{document}Distanceinvadedvs.non−invaded¯ is the mean Bray-Curtis distance between the microbial community of the focal pot and the five replicate non-invaded pots. Although this metric quantifies an invader’s ability to alter the whole microbial community, it is strongly correlated with an invader’s ability to increase invasion-facilitating microbes and decrease invasion-inhibiting microbes (Supplementary Fig. 3). It is more straightforward than the metrics specifically targeting invasion-related microbes and its quantification is easy to replicate following the widely used and standardized protocol in microbial research. To control the Type I error, we used rarefaction to normalize the relative number of microbial reads, which was conducted using the ‘rarify’ function from the R package ‘vegan’95. The Bray-Curtis distance was quantified using the ‘vegdist’ function from the R package ‘vegan’95. Because the rarefaction process produces slightly different microbial abundance matrices each time, we repeated this process 999 times and used the mean Bray-Curtis distance across the 999 microbial matrices as the MSA measure. Unlike CI, a higher MSA value indicates a stronger capacity to alter microbial communities. Therefore, we quantified the cultivator score by translating MSAs onto an axis with the minimum value set as zero and the maximum value set as 100:8\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{Cultivator}\,{score}}_{i}=\frac{{{MSA}}_{i}-\min \left({MSA}\right)}{max\left({MSA}\right)-min\left({MSA}\right)}\times 100$$\end{document}Cultivatorscorei=MSAi−minMSAmaxMSA−minMSA×100 Strategic divergence and its correlation with the non-additive effect of co-invasion Pair-wise strategic divergences between component invaders were quantified using Euclidean distances between species in a two-dimensional “competitor-cultivator” coordinate space. Each invasive species was assigned coordinates based on their competitor (x-axis) and cultivator (y-axis) scores (e.g., Ageratum: x = 57.1, y = 23.3; Chromolaena: x = 17.0, y = 56.5; Sphagneticola: x = 23.6, y = 18.7). For each co-invader combination, we quantified the overall strategic divergence as the perimeter of the triangle defined by its three constituent invaders, which sums the lengths of all three sides of the triangle. The equations for these calculations are as follows:9\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${Strategic}\,{divergence}\,{level}={\lambda }_{i{nv}1-2}+{\lambda }_{i{nv}2-3}+{\lambda }_{i{nv}1-3}$$\end{document}Strategicdivergencelevel=λinv1−2+λinv2−3+λinv1−310\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\lambda }_{i{nv}1-2}=\sqrt{{({x}_{{inv}1}-{x}_{i{nv}2})}^{2}+{({y}_{i{nv}1}-{y}_{i{nv}2})}^{2}}$$\end{document}λinv1−2=(xinv1−xinv2)2+(yinv1−yinv2)211\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\lambda }_{i{nv}2-3}=\sqrt{{({x}_{{inv}2}-{x}_{i{nv}3})}^{2}+{({y}_{i{nv}2}-{y}_{i{nv}3})}^{2}}$$\end{document}λinv2−3=(xinv2−xinv3)2+(yinv2−yinv3)212\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\lambda }_{i{nv}1-3}=\sqrt{{({x}_{{inv}1}-{x}_{i{nv}3})}^{2}+{({y}_{i{nv}1}-{y}_{i{nv}3})}^{2}}$$\end{document}λinv1−3=(xinv1−xinv3)2+(yinv1−yinv3)2 Here, λinv represents the distance between one of the three pairs of constituent invaders within a species combination. x and y represent the competitor and cultivator scores, respectively. We used simple linear regressions and Pearson correlation analyses to explore the correlation between the strength of strategic divergence and the NAE of co-invasion. This was performed using the base R package ‘stats’93. Microbial drivers We considered two microbial mechanisms driving invasional facilitations among co-invaders: (1) enrichment of invasion-facilitating taxa, or (2) depletion of invasion-inhibiting taxa. To identify key phyla involved, we implemented a seven-step framework: (1) identify invasion-facilitating candidate genera with significantly positive Pearson correlations (Pearson’s R > 0, P < 0.05) between the abundance of each genera (measured by transcripts per million) and invader biomass or competition index during single-species invasions; (2) identify invasion-inhibiting candidate genera with significantly negative Pearson correlations (Pearson’s R < 0, P < 0.05) between the abundance of each genera and invader biomass or competition index during single-species invasions; (3) identify the invasion-facilitating and invasion-inhibiting candidates associated with the constituent invaders for each co-invasion combination; (4) identify the invasion-facilitating candidates that increased and the invasion-inhibiting candidates that decreased during either single-species invasions by at least one constituent invader or co-invasions involving all three co-invaders for each combination using the one-sample t test (two-tailed); (5) quantified the fraction of invasion-facilitating microbes that simultaneously increased and invasion-inhibiting microbes that simultaneously decreased during both single-sepcies invasions and co-invasions for each combination using the Venn diagram approach; (6) identify the key microbial fraction for invasional facilitation among co-invaders through constructing linear regression models that regressed invasion outcomes (measured by non-additive effects) against the fractions of increased invasion-facilitating candidates and decreased invasion-inhibiting candidates; (7) summarized genera counts for these microbes at the phylum level and determined the phyla with the most significant changes. Correlation analyses were conducted using the R package ‘Hmisc’96. One-sample t tests and linear regressions were conducted using the base R package ‘stats’ (R Core Team 2024). Venn diagram analysis was conducted using the R package ‘ggvenn’97. Reporting summary Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article. Supplementary information Supplementary Information Description of Additional Supplementary Files Supplementary Data 1 Reporting Summary Transparent Peer Review file Supplementary Information Description of Additional Supplementary Files Supplementary Data 1 Reporting Summary Transparent Peer Review file Source data Source Data Source Data