Evolution of facilitation requires diverse communities

Abstract

Diverse experimental plant communities are more productive than monocultures. The increase of this biodiversity effect over time has been attributed to evolutionary selection for complementarity in mixtures. Here we show that evolutionary selection for enhanced net facilitative plant interactions occurred only in mixtures, while evolutionary selection for reduced net competition occurred in mixtures with mixture coexistence history and monocultures with monoculture coexistence history. Widespread declines in natural and agricultural biodiversity could therefore compromise potential evolution of facilitative interactions, that is, cornerstone processes in nature conservation and the development of sustainable agriculture.

Main

Higher species diversity increases plant primary productivity^1,2. In stable systems, where plant species are able to co-evolve, such diversity effects increase over time due to increasing species complementarity^3,4,5,6,7 resulting from evolutionary changes in species traits⁶. Greater evolved complementarity could occur through increased niche differentiation, which reduces competition intensity, or increased facilitation². Understanding the circumstances under which these two ecological processes operate and evolve is widely and substantially important. For example, it may be of benefit in the management of biodiversity–ecosystem functioning relationships, such as current efforts to develop sustainable agricultural systems through mixture cropping^8,9,10. However, we do not know whether enhanced complementarity effects in more-diverse plant communities are due to niche differentiation or facilitation^2,11, or the extent to which evolutionary forces exerted by biodiversity can influence competition and facilitation over time.

The evolutionary trajectory of competition in populations may vary between monocultures and mixtures^12,13,14. The greater potential of plants in mixtures for competitive asymmetries and the greater scale of potential niche differentiation (with two species being able to have greater divergence in traits than two genotypes) suggests evolutionary responses enabling reduced competition intensity are more likely in mixtures. Furthermore, we expect that higher competition intensity through increased niche overlap in monocultures compared to mixtures may not necessarily translate into increased selective pressure for niche differentiation and reduced competition, but rather for fitness equality¹⁵ and consequently competitive combining ability¹⁶. We therefore hypothesize that evolutionary selection for reduced competition will be stronger in mixtures than in monocultures.

The biotic agents selecting for plant facilitation are very poorly understood¹⁷. Under stable environmental conditions, the evolution of facilitation is probably dependent on neighbour identity^13,18. Beyond this, however, we have no clear basis for predicting whether intraspecific facilitation is stronger or weaker than interspecific facilitation. We therefore hypothesized that facilitation is equally likely to evolve in mixtures and in monocultures.

To test our two hypotheses, we quantified plant–plant interactions in a biodiversity–ecosystem function experiment performed under favourable growth conditions in a glasshouse with plants of monoculture, monofunctional group mixture and four-functional-group mixture coexistence histories. We quantified the intensity of negative and positive plant interactions and their dependence on historical and planted species and functional group richness and composition.

For the experiment, cuttings and seeds of 12 grassland species from 4 functional groups (legumes, grasses, small herbs and tall herbs) were collected from communities with monoculture, monofunctional group mixture and four-functional-group mixture coexistence histories and planted as single plants, monocultures, monofunctional group mixtures and multifunctional group mixtures (Fig. 1). The plants were planted in pots consisting of either one single plant (control plants) or four plants, that is, two target individuals of one species and either two neighbour individuals of the same species (monoculture communities) or two neighbour individuals of a different species (mixture communities). All plants had eight years of coexistence history of the source populations in the Jena Experiment, Germany. For the plant–plant interaction intensity assessment we calculated the cumulative neighbour-effect intensity index¹⁹ based on aboveground biomass data of the single (control) plants, and the mean biomass of the two target individuals and the two neighbour individuals in both the monoculture and mixture pots (treatments). Mortality was considered by including zero biomass for dead individuals.

Fig. 1: Experimental design.

Twelve species in the Jena Experiment were chosen from all four functional groups: grass, small herb, tall herb and legume. Plant material was collected in 2010, shoots and roots (n = 4,900), from four-functional-group mixtures (at least four species of four different functional groups), monofunctional group mixtures (at least four species of the same functional group) and monoculture (one species) communities established in Jena, Germany, in 2002. Cuttings and seedlings propagated from the plants in 2010 were assembled in monoculture, mono- or multifunctional group experimental pots in 2010–2011. Each pot was planted with a single plant, four plant multifunctional two species mixtures, four plant monofunctional two species mixtures or four plant monocultures from each of the three Jena coexistence histories. The three shades of green represent evolutionary selection in four-functional-group mixtures, monofunctional group mixtures or monoculture communities from 2002 to 2010 in the Jena Experiment.

Perhaps unsurprisingly for a mesic grassland community growing in a greenhouse, net competition (1,584 out of 1,709 plants; 92.7%) prevailed over net facilitation (125 out of 1,709 plants; 7.3%). The level of competition was dependent on coexistence history, and tended to be reduced (less negative) in planted monocultures and mixtures if the component plants had monoculture and mixture coexistence histories, respectively (Fig. 2b; Tables 1 and 2). By contrast, facilitation, which was more frequent in mixtures (112 out of 1,374 plants grown in mixtures; 8.2%) than in monocultures (13 out of 336 plants grown in monocultures; 3.9%), was more intense (more positive) in communities of plants with a mixture coexistence history compared with those with a monoculture coexistence history, without significant differences among the planted diversities (Fig. 2c; Tables 1 and 2). Importantly, these results could not be attributed only to a coexistence history with legumes or planted communities containing legumes (Supplementary Information).

Fig. 2: Less intense competition in monocultures and mixtures of respective coexistence histories, and more intense facilitation in communities with mixture history.

a, NIntC_net in monoculture and mixture communities of plants with monoculture and mixture coexistence history (n = 1,709). b,c, Differences in competitive (b) and facilitative (c) interaction intensities in monocultures and mixtures of plants from monoculture and mixture histories. In the top panels symbols are means ± s.e.m. calculated from raw data, in the bottom panels symbols are observed values. n per treatment level is shown in parentheses.

Table 1 Response of NIntC_net, NIntC_competition and NIntC_facilitation to coexistence history and planted diversity treatments

Table 2 Response of NIntC_net, NIntC_competition and NIntC_facilitation to random terms

Overall, net interactions (the combined effect of facilitation and competition) of plants in mixtures were less negative compared to plants in monocultures (F_(1,41.3) = 4.81, P = 0.034), with no difference between monofunctional and multifunctional group mixtures (F_(1,69.4) = 0.06, P = 0.806). Specifically, mixtures composed of plants with a mixture coexistence history showed fewer competitive net interactions compared to mixtures planted with plants from a monoculture coexistence history or monocultures planted with plants from a mixture coexistence history (Fig. 2a, Tables 1 and 2). This was due to both reduced competition intensity and increased facilitation intensity (Fig. 2b,c) when compared to mixtures planted with plants from a monoculture coexistence history and due to reduced competition only (Fig. 2b,c) if compared to monocultures planted with plants from a mixture coexistence history.

In addition to the coexistence history and planted diversity effects on plant–plant interaction intensities, neighbouring legumes reduced the intensity of competitive interactions in mixtures (from NIntC_competition = −0.64 ± 0.01 (mean ± s.e.m.) in the absence of neighbouring legumes to NIntC_competition = −0.54 ± 0.02 in the presence of neighbouring legumes), but not in monocultures (NIntC_competition = −0.61 ± 0.01 in the absence of neighbouring legumes to NIntC_competition = −0.76 ± 0.03 in the presence of neighbouring legumes; Tables 1 and 2). This is not surprising: the benefit of a legume neighbour will clearly be stronger for plants that lack the legumes’ ability to fix atmospheric nitrogen. What is perhaps more surprising, given that we often assume legumes are benefactors rather than beneficiaries, is that target legumes received more facilitation than target non-legumes (NIntC_facilitation = 0.30 ± 0.03 if the target was a legume and NIntC_facilitation = 0.17 ± 0.02 if the target species was a non-legume; Table 1 and 2). In fact, our additional analyses suggest that in particular the legumes’ trait of being a facilitation beneficiary might be evolutionarily selected in mixed communities (Supplementary Table 3 and Supplementary Fig. 2).

Comparing the effects of legumes, known to play a key role in plant–plant interactions and biodiversity effects on plant productivity^2,20, to the effects of historical and current plant diversity on NIntC_net, showed that the diversity effect (up to 0.081) was of similar magnitude to the effect of having legume neighbours (0.108).

Overall, our study demonstrates that whereas evolutionary selection for reduced competition intensity enhanced yield in both monocultures and mixtures, increasing yields in mixtures were also due to evolutionary selection for facilitation. There is clear evidence that plant species diversity acts as an evolutionary selective agent not only for local adaptation to the abiotic environment²¹ but also for plant–plant interactions, with effect sizes after only eight years of particular coexistence histories being comparable to those of the presence of nitrogen-fixing legumes.

It is known that complementarity among species in mixtures is produced by species sorting and evolutionary shifts in niche space^6,7,18,22,23. In our specific case, recent research has actually shown that it is the evolutionary selection of genetic variants that is responsible for the emergence of monoculture and mixture types in the Jena Experiment²⁴. Our study additionally demonstrates that facilitation and competition respond differently to this sorting out of genetic variants within plant communities of high diversity. Monocultures and mixtures composed of plants with a corresponding coexistence history showed equivalent reductions in negative net neighbour effects. However, although a coexistence history in monocultures selected for reduced facilitation, thereby counterbalancing the positive effects on yield of reduced competition in monocultures, a coexistence history in mixtures selected for increased facilitation, thereby reinforcing the increase in mixture yields over time.

In light of the current decline in natural and agricultural biodiversity^25,26,27, these findings have clear and important consequences for environmental and agricultural management. Given the importance of facilitation for species conservation²⁸, the conservation of biodiverse communities rather than restoration by replanting may be more crucial than previously thought for optimal ecosystem functioning. Our results suggest that continued monoculture cropping and artificial selection (breeding) for monocultures can indeed reduce competition intensity in agroecosystems by selection for intraspecific niche differentiation, thereby increasing intraspecific complementarity and yield. However, mixture cropping may further increase yields through potential additional facilitative interactions⁸. Our results provide clear evidence that the yield benefits of mixture cropping will only develop to their full potential if artificial selection targeted specifically for mixture crops is implemented into breeding programmes. Mixed cropping using plants with mixture coexistence histories has the potential to not only increase niche differentiation but also enable the evolution of facilitative interactions, thereby maximizing complementarity effects.

Methods

Data

We used published⁶ and new data from a biodiversity experiment conducted at the University of Zurich with 12 grassland species from the Jena Experiment (see Supplementary Data 1 for full dataset). Plant cuttings were collected from monoculture, monofunctional group mixture plots (at species diversities of 4, 8 and 16 species) and four-functional-group mixture plots (including grasses, small herbs, tall herbs and legumes at diversities of 4, 8, 16 or 60 species) of the Jena Experiment (http://www.the-jena-experiment.de). These were used to set up 1 m² plots in an experimental garden at the University of Zurich with identical plant compositions to those Jena plots from which the cuttings were collected. The plants collected in Jena had, at the time the cuttings were collected in 2010, an eight-year coexistence history since the establishment of the Jena Experiment in 2002, in either a monoculture, monofunctional group mixture or a four-functional-group mixture. The mixtures replanted in the experimental garden in Zurich were then used to propagate further cuttings (in 2010) and seeds (in 2010/2011) for the biodiversity experiment.

The biodiversity experiment was set up in a greenhouse and consisted of 12 single plants, 12 monocultures, 12 monofunctional group mixtures and 36 multifunctional group mixtures in pots planted with seedlings or cuttings with monoculture, monofunctional group mixture and four-functional-group mixture coexistence histories. Further information regarding the experimental design has been published previously⁶. Single plant pots of each species and coexistence history consisted of a single plant individual. Monoculture and mixture pots consisted of four plants per pot, with all four individuals of the same species (that is, two individuals as targets and two individuals as neighbours) and coexistence history in monoculture pots, and with two individuals per species of two species (that is one species as the target and the other species as the neighbour) of the same coexistence history in monofunctional and multifunctional group mixture pots. Each composition of single plants, monocultures, monofunctional group mixtures and multifunctional group mixtures were replicated three times, if possible⁶. The experimental set-up was replicated twice, in two blocks: one block in November 2010 with cuttings and one block in October 2011 with seeds. Cuttings and seeds from within the three different coexistence histories were randomly selected for planting as single plants, monocultures, monofunctional group mixtures or multifunctional group mixtures. Availability of cuttings and seedlings precluded some of the coexistence history × planted diversity combinations within each block. In total, 168 pots with individual plants, 168 pots with monocultures, 171 pots with monofunctional group mixtures and 516 pots with multifunctional group mixtures were grown. Plants were harvested 20 weeks after planting and aboveground dry biomass was determined for each species per pot.

Plant interaction analyses

To calculate plant–plant interactions for each interacting partner (target and neighbour) in each monoculture and mixture community we used the plants grown in isolation as controls. The average biomass of the control plant per species over all three coexistence histories allowed us to determine the direction and intensity of plant–plant interactions in monoculture and mixture communities of different coexistence histories. Specifically, we compared the performance of isolated plants growing without plant interactions in control pots to the performance of individuals growing with intra- and/or interspecific interactions with their neighbours in a monoculture or mixture community. In technical terms, for each target and neighbour species (i) biomass per individual (b) was determined in control pots (c) and community pots (t) (b_(c,i) and b_(t,i) respectively). Direction and intensity of the response of species i in community t was assessed using NIntC:

$$\begin{array}{l}{\mathrm{NIntC}}_{{\mathrm{net}}(t,i)}=2\left\{ {\left( {b_{(t,i)} - b_{(c,i)}} \right){\mathrm{/}}\left[ {\left( {b_{(t,i)} + b_{(c,i)}} \right) + |\left( {b_{(t,i)} - b_{(c,i)}} \right)|} \right]} \right\}\end{array}$$

(1)

NIntC is a simple relative effect size measure with values ranging from –1 to 1. Negative NIntC values indicate net negative (competitive) interactions and positive NIntC values indicate net positive (facilitative) interactions; more extreme values indicate stronger effect sizes¹⁹. NIntC_net quantified the relative change in individual biomass of a species in a community compared to its biomass as an individual control plant, including mortality of plants by including zero biomass for all dead plants. Comparisons of NIntC_net among monocultures and mixtures and among coexistence histories indicated differences in the net outcome of plant–plant interactions on productivity among those communities. Changes in NIntC_net from monoculture to mixture, for example, indicated either changes in competition and/or facilitation with diversity.

With the subset of targets or neighbours that showed a net negative response (NIntC_net) to their neighbour (that is, neighbour or target), we could then specifically focus on changes in net competitive interactions (NIntC_competition) associated with the diversity and coexistence history treatments. Similarly, with the subset of targets or neighbours that showed net positive responses to their neighbours (that is, neighbour or target), we could investigate changes in facilitative interactions (NIntC_facilitation) associated with the diversity and coexistence history treatments. Therefore, this method allowed us to quantify changes in net facilitation versus net competition with changing community diversity and coexistence history by comparing NIntC_facilitation and NIntC_competition among monoculture and mixture communities or among different coexistence histories. Increasing (that is, less negative) NIntC_competition indicated decreasing net competition and most likely consequent increasing niche differentiation and vice versa, whereas increasing NIntC_facilitation most likely indicated increasing facilitation intensity and vice versa.

To assess the specific role of legumes in more detail, we repeated the analyses with subsets of plants with or without a legume history and with communities planted with or without legumes. The experimental set-up included 849 plants with legume coexistence history (including all plants originating form a four-functional-group mixture) and 861 plants without legume coexistence history. Planted communities included 636 plants planted together with legumes and 1,074 plants in communities planted without legumes. We report the results of these analyses in the Supplementary Information.

Statistical analyses

We used NIntC_net, NIntC_competition and NIntC_facilitation as response variables to the diversity treatments of the coexistence history and the planted communities. We applied general linear mixed-effects models using restricted maximum likelihood estimation and assessed the significance of the fixed effects using type-I ANOVA and F tests with adjusted error terms and the Satterthwaite approximation of denominator degrees of freedom. The fixed effects of the models were block (cuttings versus seeds), species richness of the coexistence history (monoculture versus mixture), functional group richness of the coexistence history (monoculture versus monofunctional group mixture versus four-functional-group mixture), species richness of the planted community (monoculture versus mixture), functional group richness of the planted community (monoculture versus monofunctional group mixture versus multifunctional group mixture), target species is a legume (yes versus no), neighbour species is a legume (yes versus no), functional group of the target species (legume versus grass versus small herb versus tall herb), functional group of the neighbour species (legume versus grass versus small herb versus tall herb) and interactions among these. Glasshouse table, target species identity, neighbour species identity and species combination were used as random terms.

All statistical analyses were performed in R version 3.4.2.

Reporting Summary

Further information on experimental design is available in the Nature Research Reporting Summary linked to this article.

Code availability

The Supplementary Code contains the complete R-code to run all the calculations and statistical analyses.

Data availability

All data analysed during this study are included in this article and its Supplementary Information.