Understanding how the environment shapes species distribution and affects patterns of biological diversity is still a challenging task, especially in species rich regions, such as the Neotropics (Hawkins et al., 2003; Gotelli et al., 2009; Brown et al., 2020). To date, information on this topic is mostly based on vertebrates and plants, and suggest that the combination of high annual temperature with a constant supply of water and spatio-topographic complexity are the main predictors of species distribution, richness, and endemism (Hawkins et al., 2003; Kreft and Jetz, 2007; Qian, 2010; Vasconcelos et al., 2019). Within the Neotropics, the Amazon and the foothills of the North-eastern Andes are examples of regions that combine these conditions, and consequently, they exhibit high levels of species richness and phylogenetic diversity in monkeys, snakes, birds, amphibians, palms, and vascular plants (Kreft and Jetz, 2007; Fenker et al., 2014; Vallejos-Garrido et al., 2017; Velazco et al., 2021). Similarly, regions such as the Biogeographic Choco, Costa Rica, and the Amazon show high levels of phylogenetic endemism (e.g., Rosauer and Jetz, 2014; López-Aguirre et al., 2019; Varzinczak et al., 2020). However, these patterns have not been deeply evaluated in Neotropical invertebrates, and particularly butterflies (Pearson and Carroll, 2001; Mullen et al., 2011).

The environment, and especially climatic niche, has also been suggested to have an effect on gene flow. For example, phylogenetic discordance in multiple loci in beetles of the genus Mesocarabus seems to be the result of hybridization between species sharing the same climatic niche (Andújar et al., 2014), while in armadillos of the genus Dasypus, asymmetric gene flow appears to be facilitated by niche conservatisms at both sides of a geographic barrier (Arteaga et al., 2011). Additionally, climatic-based selection likely plays a role in maintaining mosaic hybrid zones in Quercus oaks, where climatic heterogeneity favors the co-occurrence of parental species and their hybrids (Swenson et al., 2008; Ortego et al., 2014).

Heliconius butterflies are a diverse insect group found across southern United States, Central, and South America, where they occupy divergent habitats (Jiggins, 2017). Due to the recent radiation of this butterfly genus, species pairs have different levels of reproductive isolation, which are used as proxies for different stages of speciation (Kronforst et al., 2013; Martin et al., 2013). In total, ∼25% of Heliconius species are known to hybridize in nature (Mallet et al., 1998, 2007), but the role of abiotic variables in facilitating or hampering such hybridization has been poorly studied (Mallet et al., 1990; Rosser et al., 2014).

In this study, we combined an extensive database of occurrences of species and hybrids in Heliconius as well as environmental data to investigate: (1) how the environment shapes the distribution of Heliconius at a regional scale, (2) how the environment molds species richness, phylogenetic diversity, and phylogenetic endemism in these butterflies, and (3) whether niche similarity promotes hybridization.

We found that Heliconius butterflies display five general distribution patterns, namely: (i) wide distribution, (ii) trans-Andes, (iii) biogeographic Choco + Costa Rica, (iv) cis-Andes + Pacific of Ecuador, and (v) highland Andes. We also found that three variables (isothermality, precipitation and altitude) explain these patterns. Isothermality is a variable that quantifies how daily temperatures oscillate relative to the annual oscillations (O’Donnell and Ignizio, 2012), and its importance as one of the most explanatory variables of species distribution is not without precedent. For example, this variable explains the distribution of frugivorous bats (Chattopadhyay et al., 2019), mealybugs (Heya et al., 2020), Opiliones (Simó et al., 2014), and American monkeys (Vallejos-Garrido et al., 2017). Although all Heliconius species are strongly affected by isothermality, its effect seems to be stronger for widely distributed species and those with trans-Andean distribution. Interestingly, these species occur in regions with high and medium isothermality (>460%), that is, in regions that experience temperature changes throughout the day but keep a constant temperature throughout the year (O’Donnell and Ignizio, 2012). This suggests that these butterflies are particularly sensitive to long term changes in temperature, thus limiting their range to tropical areas.

The distribution of species occurring in the biogeographic Choco of Colombia, Costa Rica, cis-Andes and the Pacific of Ecuador is also strongly limited by precipitation. Consistently, these regions have either rainforest, monsoon, or savanna climate, and they are the Neotropical regions with the highest precipitation [precipitation in the driest month (Pdry) > 60 mm] (Beck et al., 2018). Previous studies have suggested that cloudiness and precipitation decrease flying bout duration in butterflies and, consequently, limit their dispersal (Cormont et al., 2011). Therefore, exceptionally high levels of precipitation in such regions may act as population traps, preventing butterflies from flying over longer distances and keeping them in a single region (Rosser et al., 2014). This finding agrees with previous studies in South America, where precipitation shapes the distribution of multiple vertebrates and invertebrates (Atauchi et al., 2017; Amundrud et al., 2018; Schivo et al., 2019; de Oliveira da Conceição et al., 2020).

In addition, altitude was the best predictor for the distribution of Heliconius species that can reach elevations up to 2,600 masl, which is considerably higher than the elevational range occupied by other members of the genus (<2,200) (Rosser et al., 2012). Therefore, it is likely that these highland species have morphological or physiological modifications that allow them to expand their elevational range and occupy new niches. In fact, highland Heliconius are known to have rounder wings compared to lowland species, and this has been suggested to aid them flying dense cloud forest or compensate for the lower air pressure found at higher altitudes (Montejo-Kovacevich et al., 2019). Also, comparisons among different populations of Heliconius have revealed that highland populations are less tolerant to heat (Montejo-Kovacevich et al., 2020), which may limit their distribution range.

The foothills of the eastern Andes and the Amazon basin appeared as the regions with highest Heliconius species richness, which confirms the findings of a previous study done for the subfamily Heliconiine at a higher scale (50 km) (Rosser et al., 2012). Interestingly, both of these regions are known to present unusual concentrations of contact zones and hybrid zones (i.e., suture zones) (Dasmahapatra et al., 2010; Rosser et al., 2021), which may explain the richness they exhibit. Also, altitude, isothermality, and precipitation were the variables best correlated with this metric. This may be due to the elevational gradient found at the foothills of the eastern Andes, which offers multiple ecological niches thus favoring diversification rates (Rahbek and Graves, 2001; Jetz and Rahbek, 2002; Davies et al., 2007; Keppel et al., 2016). Additionally, there are several climate-based hypotheses that seek to explain broad-scale diversity patterns, and water and energy have emerged as crucial influencers of species richness (Silva-Flores et al., 2014). In particular, the water-energy dynamics hypothesis argues that species richness increases in places where liquid water and optimal energy conditions provide the greatest capacity for biotic dynamics (Svenning et al., 2008). The Amazon and foothills of the eastern Andes are regions with near constant hot-warm temperature throughout the year and have a permanent liquid water supply (Rosser et al., 2014; Vallejos-Garrido et al., 2017) thus ensuring an optimal water-energy dynamic. The latter translates into constant availability of plants for butterflies, including host plants for immature and pollen for adults, and continual interactions between individuals, which may be correlated with the high species richness we detected.

Similar to other studies, patterns of phylogenetic diversity were similar (although not identical) to richness (Davies Jonathan and Buckley, 2011; Fenker et al., 2014; Mendoza and Arita, 2014; Guedes et al., 2018). Interestingly, areas with highest species richness got low phylogenetic diversity (Figure 2C, blue grids), which may be a consequence of the recent increase in diversification rate in Heliconius (4.5 Ma) and the consequent co-occurrence of multiple young species in the Amazon and foothills of the eastern Andes (Rosser et al., 2012; Kozak et al., 2015). In agreement with this observation, previous research in both animals and plants have found high phylogenetic diversity in the eastern Andes of Colombia, Peru, and Ecuador (Fenker et al., 2014; Mendoza and Arita, 2014; Guedes et al., 2018; Arango et al., 2021; Velazco et al., 2021).

The highest phylogenetic endemism was found in the central eastern Andes of Colombia, and this result is possibly due to the restricted range of the species Heliconius heurippa (Figure 2D, area H1). However, we cannot rule out this result as an overestimation since the phylogenetic tree that we used (Kozak et al., 2015) considers this taxon as a separate species and not as part of H. timareta (as recently hypothesized). If H. heurippa had been included within H. timareta, which has a wider distribution range, it is likely this result on phylogenetic endemism does not hold. Additionally, the pacific region of Costa Rica, Panama and Colombia show intermediate values of phylogenetic endemism that resulted from the presence of species that have reduced geographic range and are either long-branch species (e.g., Heliconius godmani) or species for which no close relatives are known (e.g., Heliconius hewitsoni) (Figure 2D, area H2 and H3, respectively). These regions were previously described as highly endemic phylogenetically for plants (Sandel et al., 2020), terrestrial mammals (Rosauer and Jetz, 2014), birds and amphibians (Daru et al., 2020a). Interestingly, there were several species that, although are considered as geographically endemic within Heliconius, exhibited low values of phylogenetic endemism. However, it is important to acknowledge that phylogenetic endemism is a concept based on linages rather than species, and thus, if an endemic species has a narrow range but it is closely related to a widespread species, its phylogenetic endemism will not necessarily be low (Rosauer et al., 2009). An example of this is Heliconius nattereri, an endemic species from Brazil’s Atlantic Forest that, despite having a narrow distribution, is sister to the widely distributed Heliconius ethilla (Figure 2D, area H4). Similarly, Heliconius atthis is restricted to the Ecuadorian and Peruvian Pacific, but it is sister to the widely distributed Heliconius hecale (Figure 2D, area H5). In our study we found that high precipitation and near constant hot-warm temperature throughout the year are strongly correlated with phylogenetic endemism, which agrees with studies that point a role for temperature in promoting endemism by reducing extinction rates and increasing population sizes in small areas (Jetz et al., 2004; Rosauer and Jetz, 2014; Varzinczak et al., 2020).

Our environmental niche analysis showed that hybridizing species do not necessarily share the same climatic space despite some of them having largely overlapping geographic distributions. This is the case of H. ethilla and H. numata, which frequently co-occur throughout their distribution, but there are some regions with an extreme climate, such as the Pacific coast of Colombia (a humid jungle) and the Colombian Magdalena valley (which has a marked precipitation gradient, being humid in the north and dry in the south), where H. ethilla but not H. numata occur (Supplementary Figure 9). This suggests that the former species has a broader climatic tolerance. We also detected differences in the environmental niche between pairs of hybridizing species that rarely overlap geographically, but when they do, they hybridize. For example, H. erato and H. himera occupy contrasting environmental niches in Ecuador (Jiggins et al., 1997), where H. himera lives in dry forests while H. erato inhabits wet forests of the Andes (Figure 5). Similarly, the hybridizing H. erato (H. e. venus) and H. chestertonii meet in an environmental transition zone between wet and dry forest in the Colombian Andes (Muñoz et al., 2010; Supplementary Figure 8).

In summary, we confirmed that, at large scales, the distribution of Heliconius, its richness, diversity, and phylogenetic endemism are mainly shaped by a combination of high annual energy (i.e., hot-warm temperature), constant water supply, and an extraordinary topographic complexity. However, species distributions are thought to result from dynamics occurring at multiple spatial scales. Therefore, including microclimate variables and ecological interactions would provide an in-depth understanding of the multiscale drivers of distribution, niche range and phylogenetic processes (Montejo-Kovacevich et al., 2020; Paz and Guarnizo, 2020). Our study confirms the richness and diversity of areas already identified in other taxa, thus strengthening the importance for their conservation as strategic hotspots of biodiversity.

The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found below: https://doi.org/10.5281/zenodo.5149294.

NR-M and CS conceived the study. NR-M and FS-R analyzed the data. NR-M and CG-Q curated the databases. NR-M, CS, CP-D, FS-R, and CG-Q collected the specimens in regions poorly sampled in Colombia. NR-M, CS, and CP-D drafted the manuscript. All authors read and approved the final version.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.