Wing color pattern evolution
During the course of their evolution, Prepona underwent drastic changes in the color pattern of both wing surfaces (Figures 2-3). These color patterns were the key features used by previous taxonomists to classify Prepona species in two genera. The former genusAgrias contained yellow, orange, blue and red species andPrepona , as delimited previously, contained mostly blue species. The former group being nested within the latter is supported by molecular and morphological data (Ortiz-Acevedo et al., 2017), while color patterns apparently transitioned rapidly from blue to red (Figure 2), resulting in Prepona containing butterflies with such strikingly different coloration patterns (Figures 1-3).
Our results suggest that there is stronger change in phenotype for red and blue RGB channels (Table 3) than for green and total RGB. The changes identified are mostly jumps in the mean and are located around the red Prepona clade (Figure 2). The analysis consistently identified the jump from blue to red in Cell 1 with no shift in the rate of evolution. Despite the fact that the red Prepona clade is so notably different in color pattern compared to the rest of the tribe’s members, the evolutionary rate of phenoptyic change after the jump remained similar to that in blue clades. These jumps are consistent with previous findings, that drastic changes in color patterns of butterfly wings and other organisms have a relatively simple genetic basis and can appear relatively fast in evolutionary time (Nadeau et al., 2016; Reed et al., 2011). Differential genetic expression in particular regions of the wing are also responsible for localized changes (Brakefield et al., 1996; Nadeau et al., 2016; Oliver et al., 2012), which might be a plausible explanation for the results we found in Cell 1. These localized changes are not always visible under analysis of the entire wing, thus, isolating those regions that are likely to change quickly allowed us to detect the jumps in phenotype.
Jumps and shifts in the blue channel are restricted to the tips of the phylogeny and might be associated with recent speciation events. For example, we identified two jumps in the blue channel of Cell 2 (Figure 2); i) in the branch leading to Prepona amydon+P. hewitsonius and ii) in P. hewitsonius . These species are phenotypically extremely diverse in forewing coloration, ranging from dark, deep blue to light, bright blue, even within subspecies (Figure S7). We also identified a change in the Cell 3 blue channel at the branch leading to Prepona philipponi (W) . Prepona philipponi (W) has a different light blue tone compared to P. philipponi (E) (Figure S8), but with only a single studied individual of the former taxon, this result needs corroborating with further samples.
A potential explanation for the dramatic change of color patterns in Preponini is involvement in mimicry rings with genera such asCallicore and Asterope (Nymphalidae, Biblidinae), which have remarkably similar color patterns on both wing surfaces (Descimon, 1977; Jenkins, 1987; Figure S9). The inferred origin of the red clade in Preponapartially supports this hypothesis, since both Callicore andAsterope are predominantly Andean and Amazonian groups. In particular, the red Prepona showed no change in the Amazonian ancestral range as it underwent speciation (Figure 3). Mimicry is plausible since species in both groups feed on toxic plants, including the families Sapindaceae and Erythroxylaceae, but whether they sequester plant toxins as caterpillars and retain them after emerging as adults is still unknown, as is their palatability. Preliminary geographical distribution data show a correspondence in coloration and distribution among Prepona , Callicore and Asterope(Ortiz-Acevedo et al., in prep.).
Diversification of Preponines
In studies of other Neotropical butterfly groups, natural history traits and paleo-climatic events have been shown to shape diversification rates by influencing speciation and/or diversification (Chazot et al., 2020; De-Silva et al., 2016; Sahoo et al., 2017). This seems to be the case for Preponini as well. Our likelihood-based diversification rate analyses suggest a constant diversification rate over 30 Ma and is mostly congruent among alternative methods (Appendix). The diversification rate estimated by our likelihood analysis (0.12) is within the range of diversification rates estimated for other butterfly groups (Chazot et al., 2020; Peña and Espeland, 2015). However, we found support for exponentially varying speciation rates, similar to speciation rates in Neotropical Nymphalidae (Chazot et al., 2020). Extinction in the three top candidate models wasa priori fixed to zero, as estimated for larger Neotropical butterfly clades during the same geological time (Chazot et al., 2020).
Our analyses support an exponential increase of speciation promoted by the Andean uplift (Table 1). In fact, this biogeographically dependent model has slightly higher support than the time dependent models discussed above. A similar result has been found in many other organisms, both lowland and highland, and is hypothesized to result from the ecological opportunity created by emergence of new niches and the restriction of dispersal across the Andes (Smith et al 2014). Our ancestral range estimates show that, after the two last pulses in Andean uplift, lineages mostly contracted in their ranges leading to allopatric sister clades. For example, the ancestor of the clade containingArchaeoprepona licomedes , A. priene and A. chromusmost likely had an eastern Andean distribution, while its sister clade had a probable Central American ancestor (Figure 3). Also, the clade containing all Prepona species except P. dexamenusoriginated from a Chocoan ancestor, and one daughter lineage retained its ancestral range while the other dispersed to the Amazon, Guianas and Atlantic forest (Figure 3). The origin of many recent Preponini species dates to a time of intensified Andean uplift, in which the Pebas system started to drain and the Amazon river reached its current configuration, aided by the formation of the Acre system (Hoorn et al., 2010). The Miocene-Pliocene boundary was also characterized by a global vegetation change (Cerling et al., 1997).
A recent study showed that phylogenetic trees alone do not have enough information to tease apart different evolutionary scenarios since particular topology can either result from high speciation and constant extinction or decreased extinction and constant speciation (Louca and Pennell, 2020). Consequently, we would expect similar support for those alternative scenarios. Here we show however that the support for models with constant speciation and no extinction or exponentially increasing speciation with no extinction is considered to be much stronger than the models with constant speciation and decreasing extinction under an evidential statistics framework (Taper and Ponciano, 2016).
Origin and Biogeographical Patterns
Although the high Amazonian species richness of the tribe would suggest an out-of-the-Amazon biogeographical model, the tribe Preponini originated from a widespread South American ancestor at ~27.5 Ma in the early Oligocene. By this time, South America, dominated by forests (Strömberg et al., 2013), detached from Antarctica and moved north towards North America (Axelrod et al., 1991). This dating is consistent with other studies that used independent datasets (Peña and Wahlberg, 2008; Wahlberg et al., 2009). Members of Preponini likely dispersed, colonized and diverged in new niches in Central America as the bridge between land masses became more continuous at ~23 Ma (Bacon et al., 2015; Montes et al., 2015). This event has been demonstrated to have played a major role in the diversification of multiple groups of organisms (Bacon et al., 2015).
It has been suggested that some of methods for ancestral range reconstruction are biased towards estimating widespread ancestral ranges (Buerki et al., 2011; Clark et al., 2008; Matzke, 2014; Ree and Smith, 2008), as we found in the ancestor of Preponini. To reduce this bias, we restricted the ancestral estimate to contain only six of the eight areas considered, as observed in the most widespread extant Preponini species (Ronquist and Sanmartin, 2011, i.e.Archaeoprepona amphiachus, A. demophoon, A. demophon ). In addition, climatic changes accompanying the evolution of Preponini make biological sense in light of our inferred ancestral range estimate, as we discuss below.
Early speciation events in the tribe are characterized by range maintenance (~13 Ma; Nodes II and III; Figure 3).Subsequently, the ancestor of Archaeoprepona contracted in range, becoming restricted to the Atlantic forest region and subsequently dispersing to the Chocó. The transition from Atlantic Forest to Chocó happened through an anagenetic expansion including all South American areas except Western Andes and a subsequent range contraction (Table S15). The restriction of the ancestor of Archaeoprepona to Atlantic forest is coincident with a period in which tropical forests in South America were likely reduced by a decrease in global temperature (Kürschner et al., 2008; Pound et al., 2011). The late Miocene was probably characterized by grassland habitats, as suggested by high diversification of hoofed animals and changes in their dentition (Kürschner et al., 2008). Preponini species are not currently known to inhabit these grassland habitats. The subsequent dispersal to the Chocó follows an increase in global temperature and the reappearance of widespread tropical forests (Kürschner et al., 2008; Pound et al., 2011).
After the colonization of Chocó tropical forests, Archaeopreoponadispersed to central America at ~8 Ma during periods of intensified biological migration between Central and South America (Bacon et al 2015). From this point forward, high geologic activity in South America during the Pliocene (Hoorn et al., 2010) likely promoted dynamism in current ranges in which some species became widespread (clades containing Archaeoprepona demophoon and A. amphimachus ), while other became restricted (e.g A. chromus andA. priene ) to small geographic areas.
Conversely, Prepona +Mesoprepona , which originated from a South American ancestor, shows contrasting patterns of range evolution.Mesoprepona , a monotypic genus, shows a reduction of its distribution range and became restricted to eastern South America. In contrast, the relatively species-rich Prepona shows early diverging clades to have dispersed to a larger area, including South and Central America, around a time of increased migration between these two continents (Bacon et al., 2015). The early branching Prepona laertes clade was found to be initially restricted to eastern South America, with subsequent dynamism in range contraction/expansion as it underwent speciation events. More recent clades were found to have contracted their ranges at ~7 Ma, becoming restricted to the Chocó region, while the Amazon region suffered landscape reconfiguration resulting from the transition of the Pebas system to the Acre system.
The subsequent speciation events show: i) range maintenance in the restricted Prepona werneri , ii) range expansion followed by contraction in the P. pylene and P. deiphile clade, and iii) slight range expansion followed by contraction in the redPrepona clade (Node III, Figure 3). The dynamic range evolution in these clades happened in the early Pliocene ~5 Ma, which is characterized by landscape reconfiguration due to a strong activity in Andean uplift (Hoorn et al., 2010; Figure 3). This intensified final uplift of the Andes likely allowed the earliest divergent red Prepona to shift from lowland-highland distribution and become restricted to higher elevation habitats, while the ancestor of the remaining red Prepona species became restricted to the Amazonian lowlands. Current distribution ranges of the taxa in this clade show expansion of their ranges to eastern South America and dispersal to habitats west of the Andes.
Other examples of the influence of the Andean uplift on speciation and range dynamics of Prepona clades is exemplified in the clade containing P. pylene, P. eugenes, P. gnorima and P. deiphile. Recent phylogenetic reconstructions and morphological analyses of the tribe split the previously widespread Prepona pylene (sensu Lamas, 2004) intoP. pylene, P. eugenes and P. gnorima . Our dating suggests that Prepona gnorima, distributed in Central America and Chocó, split from the eastern South American P. pylene +P. eugenesat ~5 Ma. Furthermore, Prepona deiphile , although currently considered a widespread species, is paraphyletic in our phylogeny, supporting a split into a Central American and South American clade(Ortiz-Acevedo et al., 2017; Turrent Carriles and García Días, 2019) that probably happened during the last pulse of Andean uplift.
Preponini genera exhibit contrasting biogeographical histories, suggesting that they might have evolved under different evolutionary pressures and scenarios. Differences in natural history, and in particular larval host plant relationships, potentially underlie these differing biogeographical patterns as demonstrated in the closely related Neotropical charaxine tribe Anaeini for instance (Toussaint et al., 2019). Unfortunately, knowledge of host plants in Preponini is still rather incomplete, although, researchers, collectors and enthusiasts are working together to fill this gap in knowledge.
Conclusions
This first investigation of color pattern evolution in Preponini posits new hypotheses for the observed shifts across the phylogeny of the group. We found that, contrary to what might have been expected, changes in wing color did not influence diversification rates in this group. Both the formation of the Isthmus of Panama and the uplift of the Andes mountain range positively impacted the diversification of Preponini. Landscape reconfiguration and ecological opportunity created by the complex emergence of the connection between Central and South America and mountain building allowed dispersal, colonization and divergence in newly available niches. Further studies should focus on the mechanisms that triggered the striking change in color that happened rapidly in the genus Prepona . In particular, Prepona are potentially involved in mimicry rings with Callicore and Asterope , which also show a high richness in the Amazon region, with a number of species restricted to this area.