Abstract The complexity–stability debate in ecology remains unresolved in part because its empirical basis is limited. Most evidence for the predicted decline of connectance with species richness comes from food webs, leaving unclear whether this pattern extends across the full spectrum of ecological interactions. Moreover, existing results remain conceptually unresolved: connectance decreases with diversity, yet both the total number of interactions and the number of interactions per species increase. Here, we analyze 1,500 ecological interaction networks spanning diverse habitats and interaction types. We show that these patterns are broadly shared across ecological interaction networks and can be interpreted through a recent theory of information dynamics in complex networks, in which sparsity is favored by a trade-off between signal propagation and response diversity. Our results suggest that the structural component of the debate may indeed reflect a general architectural regularity of ecological communities rather than a contradiction between theory and nature.

Amazonia harbours more than 10% of the terrestrial biodiversity of the Earth1 and more than 400 Indigenous groups2. So far, however, no study has assessed how climate change and the loss of Indigenous languages may simultaneously impact its biological and cultural heritage. Here, to bridge this gap, we first assembled a database of 90,536 reports from 700 references to understand the societal benefits that native plants provide across all countries of the Amazon basin. We found that humans utilize 5,796 native plant species, which amounts to one-third of the known Amazon vascular seed plant flora. Next, analysing 8,429 species distribution models across three future climate scenarios (SSP1-2.6, SSP3-7.0 and SSP5-8.5), we show that climate change will produce a greater reduction in the ranges of utilized than of non-utilized species by 2060-2080. Locally, Indigenous cultures may lose an average of 28-34% of their utilized plant species and 18-23% of their associated services from climate change. Regionally, the loss of threatened Indigenous languages may result in a 26% reduction in the Amazonian knowledge pool. Overall, our results point to the strong climate and language vulnerability of Amazonian biocultural heritage. At the same time, these results-together with our publicly available dataset-may serve to guide biocultural restoration and reverse the growing global change effects on ecosystems and cultural traditions.

Habitat loss and fragmentation threaten biodiversity by reducing species richness and disrupting ecological interactions. But it is poorly understood how reduced habitat area and increased isolation impact community persistence, as measured by its capacity to maintain multispecies coexistence. We draw on an intensive, multi-year field study of mutualistic interactions in plant-frugivore and plant-pollinator networks across 41 islands in an insular fragmented landscape formed by dam construction in 1959. We show that the loss of island area after inundation is the primary driver that reduces the persistence of mutualistic assemblages, beyond merely reducing species richness. We further identify structural mechanisms of persistence: on larger islands, decreasing network modularity enhances persistence in both plant-frugivore and plant-pollinator communities, whereas increasing nestedness contributes to persistence only in plant-pollinator communities. These findings represent a conceptual advance in understanding the impact of habitat loss on biodiversity by showing that species loss in small habitat fragments may result from the reduced capacity of their mutualistic communities to support species coexistence, mediated by changes in network structure. We urge to include evaluations of community persistence into the design of conservation and habitat restoration strategies to more effectively mitigate the long-term impacts of habitat loss and fragmentation on biodiversity. Habitat loss can disrupt mutualistic interactions that sustain biodiversity. This study shows that larger habitat islands support more persistent plant-frugivore and plant-pollinator assemblages, linked to network modularity and nestedness.

Habitat loss poses a major threat to biodiversity. Its effects on ecological communities depend on the complex interplay between the landscape configuration-the pattern of connections between habitat patches, the pattern of interactions between species, and habitat loss patterns. Despite their individual importance, their joint effect on species persistence remains poorly understood. We explore how these three factors influence the persistence of empirical mutualistic communities. By employing spatially explicit metacommunity models, we find that landscapes with a heterogeneous distribution of connections between habitat patches exhibit high persistence under spatially uncorrelated habitat loss but are highly vulnerable to spatially correlated loss, where adjacent habitat patches are destroyed sequentially. Homogeneous landscapes with regularly arranged patches have lower persistence than heterogeneous landscapes but are more robust to correlated habitat loss. The nested structure of metacommunities enhances species persistence, with varying magnitude depending on landscape configuration and the patterns of habitat loss. These findings can help guide conservation strategies by identifying landscape and community features that promote species persistence.

Coextinctions may exacerbate the current biodiversity crisis. Yet, we do not understand all the factors that shape the robustness of communities to the loss of species. Here we analyze how coevolution influences the robustness to secondary extinctions of mutualistic and exploitative communities. We find that coevolution increases robustness in mutualism but reduces it under exploitative interactions. These differences are due to coevolution altering the density of interactions in communities. Coevolution leads to densely connected mutualistic communities and sparsely connected exploitative communities. We find the magnitude of these effects depends on the strength of coevolution and the size of the community. The largest changes to the density of interactions and robustness of communities occur when coevolutionary selection is strong. Moreover, the changes to network robustness are greater for small mutualistic communities and large exploitative communities. Our results broaden our understanding of the suite of mechanisms affecting the resilience of ecological communities. These insights may inform efforts to reduce the risk of species loss in the face of global change.

Plants may benefit from more diverse communities of arbuscular mycorrhizal fungi (AMF), as functional complementarity of AMF may allow for increased resource acquisition, and because a high AMF diversity increases the probability of plants matching with an optimal AMF symbiont. We repeatedly radiolabeled plants and AMF in the glasshouse over c. 9 months to test how AMF species richness (SR) influences the exchange of plant C (14C) for AMF P (32P & 33P) and resulting shoot nutrients and mass from a biodiversity-ecosystem functioning perspective. Plant P acquisition via AMF increased with sown AMF SR, as did shoot biomass, shoot P, and shoot N. The rate of plant C transferred to AMF for this P (C:P) decreased with sown AMF SR. Plants in plant communities benefit from inoculation with a variety of AMF species via more favorable resource exchange. Surprisingly, this effect did not differ among functionally distinct communities comprised entirely of either legumes, nonlegume forbs, or C3 grasses.

AimPlant recruitment involves both stochastic and deterministic processes. Recruits may establish independently or interact nonrandomly with canopy plants. We explore this deterministic aspect by testing whether recruitment patterns are influenced by the phylogenetic history of canopy and recruiting plants. Since the effect of canopy plants in recruitment can be positive (facilitation), negative (competition) or neutral, we also estimated the phylogenetic signal separately for each interaction type. Furthermore, we assessed whether environmental stress influenced the phylogenetic signal, under the expectation that more severe environmental conditions will lead to stronger phylogenetic signatures in network structure.LocationGlobal.Time Period1998-2021.Major Taxa StudiedAngiospermae.MethodsWe analysed recruitment interactions occurring in 133 plant communities included in the RecruitNet database, which encompasses a wide range of biomes and vegetation types. The phylogenetic signal in canopy-recruit interactions was quantified in different dimensions of the recruitment niche, represented by the level of interaction generalisation, and by the taxonomic and evolutionary composition of the group of canopy plants.ResultsWe found significant phylogenetic signals in more networks than expected by chance. Canopies' evolutionary history influenced facilitative and competitive but not neutral interactions. The phylogenetic signal in the recruitment niche strengthened in arid regions, suggesting that stressful habitats promote the occurrence of conserved recruitment interactions where closely related species recruit in association with closely related canopy species.Main ConclusionsDespite the strong influence of stochastic processes on plant recruitment, evolutionary history plays a significant role in driving the recruitment process, especially in harsh environments. In particular, the historical effect becomes more important when canopy species have a significant impact on the performance of recruits, either through facilitation or competition. More generally, we show that the analysis of different dimensions of the ecological niche can reveal important insights on the functional roles of interacting species.

The Geographic Mosaic Theory of Coevolution (GMTC) predicts that reciprocal evolutionary effects vary across landscapes, generating hotspots and coldspots. Traditionally, these states are treated as discrete categories, even though the intensity of coevolutionary selection can vary continuously. To capture this variation, we introduce a concept of coevolutionary temperature, ranging from coldspot to hotspot. We propose two complementary metrics to quantify it: reciprocity and strength of pairwise evolutionary effects. We also extend the GMTC framework beyond its traditional focus on pairwise systems to species-rich communities. Applying this approach to empirical plant-pollinator networks in a fragmented landscape, we find pronounced geographic mosaics in coevolutionary temperature. Smaller habitat patches support small, highly connected, and weakly nested communities with high reciprocity and strength, suggesting that they act as coevolutionary hotspots. In contrast, larger patches host species-rich, poorly connected, and highly nested communities with low reciprocity and strength, consistent with coldspots. At the interaction scale, reciprocity depends on degree similarity, with interactions between species that have similar numbers of partners exhibiting higher reciprocity. Together, these results highlight the strong dependence of coevolutionary effects on spatial variation in community structure and show how extending the geographic mosaic framework to species-rich communities can deepen our understanding of coevolution in complex systems. ### Competing Interest Statement The authors have declared no competing interest. European Commission, https://ror.org/00k4n6c32, EP/Z000831/1 University of Zurich, FK-22-114 Instituto Serrapilheira, 1912-32354 Comunidad de Madrid, 2022-T1/AMB-24091

The origin of eukaryotes represents one of the most significant events in evolution since it allowed the posterior emergence of multicellular organisms. Yet, it remains unclear how existing regulatory mechanisms of gene activity were transformed to allow this increase in complexity. Here, we address this question by analyzing the length distribution of proteins and their corresponding genes for 6,519 species across the tree of life. We find a scale-invariant relationship between gene mean length and variance maintained across the entire evolutionary history. Using a simple model, we show that this scale-invariant relationship naturally originates through a simple multiplicative process of gene growth. During the first phase of this process, corresponding to prokaryotes, protein length follows gene growth. At the onset of the eukaryotic cell, however, mean protein length stabilizes around 500 amino acids. While genes continued growing at the same rate as before, this growth primarily involved noncoding sequences that complemented proteins in regulating gene activity. Our analysis indicates that this shift at the origin of the eukaryotic cell was due to an algorithmic phase transition equivalent to that of certain search algorithms triggered by the constraints in finding increasingly larger proteins.

Landscape-scale ecological restoration is a key strategy for halting and reversing biodiversity decline. However, ensuring the long-term sustainability of restoration efforts requires guiding the recovery of complex ecological systems with many interdependent species at a landscape scale. Due to these challenges, our understanding of recovery trajectories remains limited. Using metacommunity models and experiments, we explore how the spatial configuration of communities and food-web complexity jointly influence species recovery at different spatial scales. We find that the number and spatial placement of communities affect the colonisation of empty habitat patches, but do not influence population recovery in patches where communities are introduced. Food-web complexity reduces the recovery of lower trophic levels. However, this negative effect may be partially mitigated at higher levels of food-web complexity. Our results demonstrate that the joint consideration of spatial configuration and species interactions could enhance the effectiveness of restoration actions.

Ecosystems can undergo abrupt critical transitions even when environmental change is gradual, resulting in hysteresis, where recovery requires conditions far more favorable than those that triggered collapse. While previous research has mainly focused on local network dynamics, the role of spatial heterogeneity and dispersal in mutualistic ecosystem resilience remains less explored. This study examines how spatial structures—grid, random, small-world, and scale-free networks—interact with dispersal rates to influence ecosystem recovery and mutualistic network persistence. We find that with low dispersal, restoration cost is similar across spatial structures, but at intermediate dispersal rates, scale-free networks show faster recovery and smaller restoration costs. At high dispersal rates, increased connectivity initially reduces restoration cost; however, over time, homogenization weakens spatial heterogeneity, causing restoration cost to increase. Importantly, increasing nestedness can delay collapse but also extends the recovery distance, making ecosystems harder to restore. By adjusting dispersal rates and transitioning from homogeneous to heterogeneous spatial structures, we can decrease restoration cost and improve ecosystem stability, offering key insights for ecological management strategies. ### Competing Interest Statement The authors have declared no competing interest. Swiss National Science Foundation, https://ror.org/00yjd3n13, 310030 197201

Soil fungi are a key constituent of global biodiversity and play a pivotal role in agroecosystems. How arable farming affects soil fungal biogeography and whether it has a disproportional impact on rare taxa is poorly understood. Here, we used the high-resolution PacBio Sequel targeting the entire ITS region to investigate the distribution of soil fungi in 217 sites across a 3000 km gradient in Europe. We found a consistently lower diversity of fungi in arable lands than grasslands, with geographic locations significantly impacting fungal community structures. Prevalent fungal groups became even more abundant, whereas rare groups became fewer or absent in arable lands, suggesting a biotic homogenization due to arable farming. The rare fungal groups were narrowly distributed and more common in grasslands. Our findings suggest that rare soil fungi are disproportionally affected by arable farming, and sustainable farming practices should protect rare taxa and the ecosystem services they support.

The web of interactions in a community drives the coevolution of species. Yet it is unclear how the outcome of species interactions influences the coevolutionary dynamics of communities. This is a pressing matter, as changes to the outcome of interactions may become more common with human-induced global change. Here, we combine network and evolutionary theory to explore coevolutionary outcomes in communities harboring mutualistic and antagonistic interactions. We show that as the ratio of mutualistic to antagonistic interactions decreases, selection imposed by direct partners outweighs that imposed by indirect partners. This weakening of indirect effects results in communities composed of species with dissimilar traits and fast rates of adaptation. These changes are more pronounced when specialist consumers are the first species to engage in antagonistic interactions. Hence, a shift in the outcome of species interactions may reverberate across communities and alter the direction and speed of coevolution.

Biological networks are often modular. Explanations for this peculiarity either assume an adaptive advantage of a modular design such as higher robustness, or attribute it to neutral factors such as constraints underlying network assembly. Interestingly, most insights on the origin of modularity stem from models in which interactions are either determined by highly simplistic mechanisms, or have no mechanistic basis at all. Yet, empirical knowledge suggests that biological interactions are often mediated by complex structural or behavioural traits. Here, we investigate the origins of modularity using a model in which interactions are determined by potentially complex traits. Specifically, we model system elements—such as the species in an ecosystem—as finite-state machines (FSMs), and determine their interactions by means of communication between the corresponding FSMs. Using this model, we show that modularity probably emerges for free. We further find that the more modular an interaction network is, the less complex are the traits that mediate the interactions. Altogether, our results suggest that the conditions for modularity to evolve may be much broader than previously thought.

Arbuscular mycorrhizal fungi (AMF) are plant root symbionts that provide phosphorus (P) to plants in exchange for photosynthetically fixed carbon (C). Previous research has shown that plants-given a choice among AMF species-may preferentially allocate C to AMF species that provide more P. However, these investigations rested on a limited set of plant and AMF species, and it therefore remains unclear how general this phenomenon is. Here, we combined 4 plant and 6 AMF species in 24 distinct plant-AMF species compositions in split-root microcosms, manipulating the species identity of AMF in either side of the root system. Using 14C and 32P/33P radioisotope tracers, we tracked the transfer of C and P between plants and AMF, respectively. We found that when plants had a choice of AMF species, AMF species which transferred more P acquired more C. Evidence for preferential C allocation to more beneficial AMF species within individual plant roots was equivocal. However, AMF species which transferred more P to plants did so at lower C-to-P ratios, highlighting the importance both of absolute and relative costs of P acquisition from AMF. When plants had a choice of AMF species, their shoots contained a larger total amount of P at higher concentrations. Our results thus highlight the benefits of plant C choice among AMF for plant P acquisition.

Amazonia harbors one fourth of the world's plant diversity and over 300 Indigenous groups. So far, however, no study has assessed how climate change may simultaneously impact its biological and cultural heritage. To bridge this gap, we assembled a database on 5,833 utilized plant species and show that climate change will reduce more the ranges of utilized than of non-utilized species by 2070. Locally, Indigenous cultures may lose an average of 65% of their utilized plant species and 50% of their associated services from climate change. Regionally, the loss of threatened languages may result in a 41% reduction in the Amazonian knowledge pool. Overall, our results point to the strong climate vulnerability of Amazonian biocultural heritage. ### Competing Interest Statement The authors have declared no competing interest.

Plant-hummingbird interactions are considered a classic example of coevolution, a process in which mutually dependent species influence each other's evolution. Plants depend on hummingbirds for pollination, whereas hummingbirds rely on nectar for food. As a step towards understanding coevolution, this review focuses on the macroevolutionary consequences of plant-hummingbird interactions, a relatively underexplored area in the current literature. We synthesize prior studies, illustrating the origins and dynamics of hummingbird pollination across different angiosperm clades previously pollinated by insects (mostly bees), bats, and passerine birds. In some cases, the crown age of hummingbirds pre-dates the plants they pollinate. In other cases, plant groups transitioned to hummingbird pollination early in the establishment of this bird group in the Americas, with the build-up of both diversities coinciding temporally, and hence suggesting co-diversification. Determining what triggers shifts to and away from hummingbird pollination remains a major open challenge. The impact of hummingbirds on plant diversification is complex, with many tropical plant lineages experiencing increased diversification after acquiring flowers that attract hummingbirds, and others experiencing no change or even a decrease in diversification rates. This mixed evidence suggests that other extrinsic or intrinsic factors, such as local climate and isolation, are important covariables driving the diversification of plants adapted to hummingbird pollination. To guide future studies, we discuss the mechanisms and contexts under which hummingbirds, as a clade and as individual species (e.g. traits, foraging behaviour, degree of specialization), could influence plant evolution. We conclude by commenting on how macroevolutionary signals of the mutualism could relate to coevolution, highlighting the unbalanced focus on the plant side of the interaction, and advocating for the use of species-level interaction data in macroevolutionary studies.

Complex systems ranging from societies to ecological communities and power grids may be viewed as networks of connected elements. Such systems can go through critical transitions driven by an avalanche of contagious change. Here we ask, where in a complex network such a systemic shift is most likely to start. Intuitively, a central node seems the most likely source of such change. Indeed, topological studies suggest that central nodes can be the Achilles heel for attacks. We argue that the opposite is true for the class of networks in which all nodes tend to follow the state of their neighbors, a category we call two-way pull networks. In this case, a well-connected central node is an unlikely starting point of a systemic shift due to the buffering effect of connected neighbors. As a result, change is most likely to cascade through the network if it spreads first among relatively poorly connected nodes in the periphery. The probability of such initial spread is highest when the perturbation starts from intermediately connected nodes at the periphery, or more specifically, nodes with intermediate degree and relatively low closeness centrality. Our finding is consistent with empirical observations on social innovation, and may be relevant to topics as different as the sources of originality of art, collapse of financial and ecological networks and the onset of psychiatric disorders.

Theory suggests that increasingly long, negative feedback loops of many interacting species may destabilize food webs as complexity increases. Less attention has, however, been paid to the specific ways in which these delayed negative feedbacks' may affect the response of complex ecosystems to global environmental change. Here, we describe five fundamental ways in which these feedbacks might pave the way for abrupt, large-scale transitions and species losses. By combining topological and bioenergetic models, we then proceed by showing that the likelihood of such transitions increases with the number of interacting species and/or when the combined effects of stabilizing network patterns approach the minimum required for stable coexistence. Our findings thus shift the question from the classical question of what makes complex, unaltered ecosystems stable to whether the effects of, known and unknown, stabilizing food-web patterns are sufficient to prevent abrupt, large-scale transitions under global environmental change.