Warm and arid regions of the world are hotspots of superorganism complexity

1. Introduction

Phenotypic complexity has been studied in relation to ecology and evolution of lineages, but the modern-day geographical distribution of such complexity remains enigmatic [1]. Identifying such geographical hotspots of phenotypic complexity may help provide a window into the environmental conditions that promote phenotypic innovation. However, if asked which part of the world comprises an overabundance of extant and highly complex organisms, one would probably not know the answer.

The concept of phenotypic complexity can be defined in a number of ways and described at various levels of organizations: from genes, cells, individuals to societies [2]. McShea [2], for example, defines complexity as the number of different part types at a given hierarchical level. Within a multicellular organism, phenotypic complexity is most often quantified as the number of cell types [3]. A eusocial insect colony, such as that of ants, bees, wasps and termites, is thought to be analogous to a superorganism with a reproductive queen caste functioning as its germline and non-reproductive worker caste functioning as its soma [4–7]. In solitary multicellular organisms, a single genome can produce differentiated cell types in response to cues from its internal environment, like morphogen gradients. Similarly, in superorganisms, a single genome can also produce morphologically differentiated castes within a single colony in response to external, such as nutrition, and internal, such as hormonal, cues [8–11]. Therefore, eusocial insects offer a unique opportunity to study the geography and environmental drivers of superorgansim complexity.

In addition to the morphologically distinct reproductive queen caste and a non-reproductive worker caste, some species of ants display remarkable variations in the worker caste [12]. Worker ants of the same colony can differ dramatically in their size and head-to-body scaling (see figure 1; electronic supplementary material, figure S1). This interindividual variation within a colony is called ‘worker polymorphism' and is associated with enhanced division of labour [8]. In some cases, polymorphic ant species exhibit wide variation in body size with little variation in head to body scaling [13]. Such species are considered more morphologically complex than species with limited worker body size variation (i.e. monomorphic worker castes) but less complex than species with discrete morphological castes, which vary tremendously in both size and head-to-body scaling (reviewed in [5]). Therefore, superorganisms fall along a continuum of phenotypic complexity that is governed by multiple-levels of selection, from individual to colony-level selection [7,9,14]. Furthermore, previous work has identified a suite of traits that generally correlate with the morphological complexity of superorganisms, including colony size, reproductive division of labour, queen–worker dimorphism and worker polymorphism [8,12,14–18]. The degree of morphological differentiation within the worker caste of a single colony can thus be used as one proxy for the complexity of superorganisms across ants [8,19] (figure 1).

Worker polymorphism evolved at least eight times within ants [13] and was previously estimated to occur in 13% of the approximately 330 ant genera [20]. As an example, several of the genera wherein worker polymorphism evolved, such as the leaf-cutting ants (i.e. Atta), carpenter ants (i.e. Camponotus), army ants (i.e. Eciton), fire ants (i.e. Solenopsis) and big-headed ants (e.g. Pheidole), comprise species with larger colony sizes (thousands to millions of ants within a colony), dramatic morphological differences between workers in the colony and enhanced division of labour [5]. Many of these same species have strong effects on the structure of food webs and ecosystems as dominant predators, herbivores and ecosystem engineers [21]. In addition, morphologically diverse ant species have often been argued to have enhanced cooperation and division of labour and a higher propensity for ecological and evolutionary dominance [12,13,22–24]. For these reasons, the ecology, evolution and development of worker polymorphism in ants have been extensively studied in a select set of species and in a few regions of the world [11,13,23,25]. Yet, it remains unclear where in the world one is more likely to encounter a polymorphic ant colony, and whether abiotic conditions are associated with such superorganism complexity. Some have suggested that polymorphism might be more common in tropical regions [26], but this hypothesis is largely anecdotal. To our knowledge, no other hypotheses predicting how worker polymorphism is geographically distributed have been proposed.

Owing to the current lack of process-based hypotheses, we aim to generate the first global map of the geography of worker polymorphism and assess the influence of the abiotic environment on this pattern. Specifically, we investigate the drivers of the global geography of ant worker polymorphism (superorganism complexity) using a dataset including 8990 ant species and 256 754 spatial occurrences. We first identify global hotspots of superorganisms complexity defined as climatic zones where the probability of encountering a worker polymorphic ant colony is greatest relative to the total number of ant colonies and ant species recorded in the region. To do this, we model each occurrence data point as representing an ant colony that is either polymorphic or monomorphic. Then, we compare model fit in a series of linear models relating the presence/absence of polymorphism in an ant colony to ambient climatic conditions. In addition, because the probability of encountering a polymorphic ant colony is highest where the total number of ant colonies (sampling effort) and species (species richness (SR)) is highest by chance alone, we ran a series of models aimed at evaluating the influence of these covariates on model selection. We present our results and focus the discussion on models that account for differences in sampling effort and SR, which helps us identify climatic zones that are dominated by complex superorganisms.

2. Methods

3. Results

Our data show that worker polymorphism occurs in approximately 29% of all ant species, which is a much larger number of worker polymorphic ant lineages than previously thought (electronic supplementary material, tables S1 and S2) [10,20,29]. Specifically, 4732 of the 15 518 species in our database are polymorphic (i.e. ranging from monophasic to trimorphic allometry; electronic supplementary material, table S1 and figure S1), which is double the previous estimate of 13% [20,29]. Using a threshold of greater than 50% of species within the genus being polymorphic, we found that 85 (25%) of the 337 ant genera are polymorphic (electronic supplementary material, table S2). Taken together, our results indicate that many lineages possess the developmental capacity to generate substantial variation in size and/or head-to-body scaling between individuals within a single colony [20,45,46].

To estimate the probability of worker polymorphism, we assessed the relationship between the probability of polymorphism and climate. We discovered that the probability of polymorphism is highest in arid regions of the world (figure 2). Specifically, the model with the lowest AIC was one including a positive linear term for temperature, a negative linear term for precipitation and a quadratic term for precipitation (tables 1 and 2; electronic supplementary material, table S4). The probability of encountering a polymorphic species is therefore highest in the warmest regions and very dry regions of the world (electronic supplementary material, figures S3 and S5). We also conducted sensitivity analyses to assess the influence of (i) sampling effort (SiteID as random effect) and (ii) SR (as fixed effect) on model selection (electronic supplementary material, tables S4 and S5) and (iii) the influence of using an equal area grid cell projection on our model selection (electronic supplementary material, table S6). Our results were robust to all sensitivity analyses. Finally, as a post hoc analysis, we used the model with the random effect to predict the probability of polymorphism among climatic zones (i.e. biomes) of the world (figure 3). Using this model, the probability of encountering a polymorphic ant colony is highest in arid regions around the world including deserts and tropical savannahs (figure 3).

Finally, we assessed the potential influence of evolutionary history on the global distribution of polymorphism. We first compared model fit separately for each subfamily. We found that seven of the eight subfamilies show that polymorphism is positively related with temperature, while for five of the eight subfamilies, it is negatively related with precipitation (table 3; electronic supplementary material, table S7). Furthermore, we quantified phylogenetic signal across 306 ant genera and found that closely related genera are no more likely to be polymorphic than distantly related ones (electronic supplementary material, table S8). This is consistent with the fact that polymorphism evolved several times independently in specific genera. Together, these analyses show that the global distribution of polymorphism is not a consequence of differences in the evolutionary history of subfamilies and genera. However, there may be an effect of the evolutionary history of species within genera. A large majority of polymorphic species are in the hyper-diverse genera Camponotus and Pheidole. However, these genera are present and diverse in almost all biomes of the world (electronic supplementary material, figure S6), suggesting that the distributions of these genera do not have a disproportionate influence on the global geography of polymorphism. The genus Cephalotes appears to be the third most diverse polymorphic genera in tropical dry forests and savannahs, while the genera Myrmecosystus, Messor and Cataglyphis comprise a large proportion of polymorphic species in deserts and scrublands. Therefore, these genera, which evolved worker polymorphism independently, contribute substantially to the peak in polymorphism in arid regions.

4. Discussion

Our results show that complex superorganisms are overrepresented in the warmest and most arid regions of the world, which typically harbour fewer ant species [47,48]. Specifically, the probability of encountering worker polymorphic colonies peak in tropical dry forests, savannahs and deserts. Although the processes underpinning the formation and maintenance of these organismic complexity hotspots remain enigmatic [8,9,11,23,49], this first outlook into the geography of ant worker polymorphism provides novel insights that will guide future research.

Arid systems emerge as hotspots of complex superorganisms (figures 2 and 3). These results show that worker polymorphic ant colonies are most common in warm and dry regions of the world. There are several hypotheses that may explain why polymorphic ants dominate arid systems. One possibility is that worker polymorphism in arid regions is not adaptive, but rather a consequence of evolutionary and biogeographic history. In this scenario, the current geographical distribution of worker polymorphism in particular lineages reflects that of their ancestors. However, our assessment of historical effects on the over-representation of worker polymorphism in arid environments suggests that these effects are negligeable. Furthermore, some of the most diversified and polymorphic ant genera in arid systems (e.g. Camponotus, Pheidole, Cephalotes, Cataglyphis, Messor) are not closely related and evolved independently, perhaps through convergent evolution mediated by strong selection pressure in these stressful environments [20,23]. It is likely that worker polymorphism enables persistence in extremely hot and resource-scarce environments. This is not to say that worker polymorphism originated in such environments but that it provides fitness advantages or permit survival in such environments.

The over-representation of polymorphic ants in arid regions of the world thus implies that phenotypic complexity enables persistence in these stressful environments. First, there might be size-related differences in the diet of ant workers within a colony. Arid ecosystems are numerically dominated by ants that specialize in the harvesting of seeds. Such diet specialization may only be possible, or optimized, if colonies can exploit a wide array of seed size and shapes [50]. Large workers of several seed harvester ants, for example, are known to specialize on larger food items than do smaller ants within the colony [51]. Added to the fact that deserts and savannahs are resource-limited and unpredictable, polymorphism might allow more efficient exploitation of specific and limited seed resources [52]. Second, there might be size-related differences in the thermal and drought tolerance of workers within ant colonies [53]. A few studies have shown that such differences exist, where larger workers within the colony forage at midday when temperature is at its peak and smaller worker forage when temperature cools down [54]. Moreover, in at least two species belonging to two different genera (i.e. Cataglyphis and Messor), size-related differences in the foraging schedule of workers correspond with higher tolerance to extreme temperature and to desiccation [54,55]. In sum, evidence for the hypothesis that worker polymorphism promotes persistence in arid environment is generally sparse and anectodical for the moment being, but our results provide additional incentive to test its generality.

Our findings further suggest that more generally, warm temperature is a key factor influencing the occurrence of complex superorganisms across the globe. One hypothesis to explain this trend is that warm temperatures promote worker polymorphism by shaping the biotic environment outside the colony. Many claim that biotic interactions are more intense in warm climates [56], which might favour the polymorphic phenotype, and in particular large workers involved in colony defence, through selection and social regulation [57,58]. Consistent with this hypothesis, ecological specialization, driven by intense resource competition, explains the evolution of polymorphism within at least two polymorphic genera [52,59,60]. Another hypothesis is that ecological factors such as diet and nutrition, which maybe more rich and abundant in warm regions, can also interact with developmental pathways and lead to the evolution of a greater diversity of phenotypes within the worker caste of an ant colony [17,61]. Worker polymorphism is a social trait mediating resource exploitation, resource monopolization and colony defence, both directly and indirectly affecting energy acquisition by a colony. In tropical environments, for example, resources are abundant, but competition is intense, which calls for efficient exploitation strategy including diet specialization and resource defence, both of which might relate to worker morphological complexity [56,62,63]. Polymorphism may have evolved to better defend space or more efficiently acquire resources [59,64].

We hope our findings will facilitate our understanding of the relative contributions of developmental processes, social behaviour and the ecological environment in shaping spatial variation in the expression of phenotypic complexity, which will be the key to elucidating the geography of organismic and super-organismic complexity [17,45,46,52,59,60,64–68]. Looking forward, we need to develop standardized and repeatable quantitative approaches to compare the degree of phenotypic complexity within and across species, which would allow stronger comparative studies in macroecology and macroevolution. As a first step, worker polymorphism can be quantified by relating head width (y-axis) to body length (x-axis) for individual workers within a single colony [45]. Rigorous tests of hypotheses regarding the distribution and evolution of phenotypic complexity could then be tested on solitary multicellular organisms as well as on superorganisms. Complexity can be defined and quantified across hierarchical levels of complexity, from genomic to cellular to phenotypic complexity. To accomplish this in eusocial insects, we also need to create comprehensive databases for other traits associated with superorganism complexity, namely colony size, reproductive division of labour and the number and degree of partitioning of behaviorial tasks [5,18]. Finally, whereas we investigated the two main axes of climatic variation on Earth (i.e. temperature and precipitation), which are typically used to define climatic zones, future work could investigate the influence of seasonality or climatic predictability, which could also play a role in the geographical distribution of polymorphism. The present study nevertheless represents an important first step in furthering our understanding of what facilitates and what restricts the global distribution of phenotypic complexity.

Finally, warm, and in particular arid, regions of the world harbour the greatest density of complex superorganism, which has intriguing implications in the context of climate change. Ongoing climate change leads to the aridification of many of the world's ecosystems, increasing the frequency and intensity of fires and drought events, which threatens biological diversity [69–72]. On the other hand, our findings suggest that ecosystem aridifications favour the survival of highly complex superorganisms, which often play crucial roles in decomposition and nutrient cycling, among other things [73]. The plasticity conferred to polymorphic species could also help them cope with changing abiotic and biotic conditions amid global changes. Moreover, if arid systems are hotspots of phenotypic complexity and potentially epicentres of morphological evolution and innovation in eusocial insects, then the biodiversity climate crisis may either promote further phenotypic innovation if they can adapt or drive our most complex superorganisms to extinction if they cannot. A look at the biodiversity crisis through the lenses of phenotypic complexity, a neglected dimension of biological diversity, could help identify and protect some of the world's hidden evolutionary centres.

Supplementary Material

Data accessibility

Data on ant species occurrence and polymorphism classification are available from the Dryad Digital Repository: https://datadryad.org/stash/share/2WYof3lmWteaWz05918NCp8ZtgUEeXYScxUTuJiSZI4 [75].

Authors' contributions

F.L.R.: conceptualization, data curation, formal analysis, methodology, writing—original draft; G.M.: formal analysis, methodology, visualization, writing—review and editing; B.G.: data curation, writing—review and editing; R.R.D.: funding acquisition, writing—review and editing; E.E.: funding acquisition, writing—review and editing; S.P.: data curation, writing—review and editing; N.J.S.: funding acquisition, writing—review and editing; M.D.W.: data curation, writing—review and editing; E.A.: conceptualization, data curation, funding acquisition, investigation, methodology, supervision, writing—original draft, writing—review and editing; J.-P.L.: conceptualization, data curation, funding acquisition, investigation, methodology, supervision, writing—original draft, writing—review and editing.

All authors gave final approval for publication and agreed to be held accountable for the work performed therein.

Competing interests

We declare we have no competing interests.

Funding

This research was supported by an NSERC Discovery Grant (RGPIN-2015-06081) and an FRQNT New Researcher Grant (2016-NC-189273) to J.-P.L., an NSERC Discovery Grant (RGPIN-2020-05175) to E.A. M.D.W. was supported by NSF (MSB-FRA no 1702426). G.M. was supported by a Concordia Graduate Fellowship and Concordia Excellence Award. E.E. was supported by subsidy funding to OIST and a grants from the Japan Ministry of the Environment (4-1904) and Japan Society for the Promotion of Science KAKENHI (JP17K15180). The Quebec Centre for Biodiversity Science supported the research and training of F.L.R. and G.M.