Diversification dynamics in the Neotropics through time, clades, and biogeographic regions

  1. Andrea S Meseguer  Is a corresponding author
  2. Alice Michel
  3. Pierre-Henri Fabre
  4. Oscar A Pérez Escobar
  5. Guillaume Chomicki
  6. Ricarda Riina
  7. Alexandre Antonelli
  8. Pierre-Olivier Antoine
  9. Frédéric Delsuc
  10. Fabien L Condamine
  1. Institut des Sciences de l’Evolution de Montpellier (Université de Montpellier | CNRS | IRD | EPHE), Place Eugène Bataillon, France
  2. Real Jardín Botánico (RJB), CSIC, Spain
  3. Department of Anthropology, University of California, United States
  4. Mammal Section, The Natural History Museum, Cromwell Road, United Kingdom
  5. Institut Universitaire de France (IUF), France
  6. Royal Botanic Gardens, United Kingdom
  7. Ecology and Evolutionary Biology, University of Sheffield, United Kingdom
  8. Gothenburg Global Biodiversity Centre, Department of Biological and Environmental Sciences, University of Gothenburg, Sweden
  9. Department of Plant Sciences, University of Oxford, United Kingdom

Abstract

The origins and evolution of the outstanding Neotropical biodiversity are a matter of intense debate. A comprehensive understanding is hindered by the lack of deep-time comparative data across wide phylogenetic and ecological contexts. Here, we quantify the prevailing diversification trajectories and drivers of Neotropical diversification in a sample of 150 phylogenies (12,512 species) of seed plants and tetrapods, and assess their variation across Neotropical regions and taxa. Analyses indicate that Neotropical diversity has mostly expanded through time (70% of the clades), while scenarios of saturated and declining diversity account for 21% and 9% of Neotropical diversity, respectively. Five biogeographic areas are identified as distinctive units of long-term Neotropical evolution, including Pan-Amazonia, the Dry Diagonal, and Bahama-Antilles. Diversification dynamics do not differ across these areas, suggesting no geographic structure in long-term Neotropical diversification. In contrast, diversification dynamics differ across taxa: plant diversity mostly expanded through time (88%), while a substantial fraction (43%) of tetrapod diversity accumulated at a slower pace or declined towards the present. These opposite evolutionary patterns may reflect different capacities for plants and tetrapods to cope with past climate changes.

Editor's evaluation

This important work by Meseguer et al. depicts findings that substantially advance our understanding of clade diversification across major Neotropical bioregions. The evidence that summarises the evolutionary diversity dynamics of 150 time-calibrated clades of neotropical plants and animals data is convincingly presented with current state-of-the-art analyses. The work will be of interest to evolutionary biologists and biogeographers working to understand the origins of the most biodiverse land mass on the planet.

https://doi.org/10.7554/eLife.74503.sa0

Introduction

Comprising most of South America, Central America, tropical Mexico, and the Caribbean Islands, the Neotropics are the most biodiverse region on Earth, home to at least a third of global biodiversity (Raven et al., 2020). This region not only includes the largest tropical rainforest, Amazonia, but also 8 of the world’s 34 biodiversity hotspots (Mittermeier et al., 2011). The tropical Andes, in particular, are considered to be the most species-rich region in the world for amphibians, birds, and plants (Myers et al., 2000), while Mesoamerica and the Caribbean Islands are the richest regions for squamates, and Amazonia has been identified as the primary biogeographic source of Neotropical biodiversity (Antonelli et al., 2018c). The drivers underlying the origins and maintenance of the extraordinary biodiversity of the Neotropics are hotly debated in evolutionary ecology and remain elusive (Gentry, 1982; Simpson, 1980; Antonelli and Sanmartín, 2011a; Hoorn et al., 2010; Rull, 2011; Antonelli et al., 2018a).

Attempts to explain Neotropical diversity traditionally relied on two evolutionary models. In the first, tropical regions are described as the ‘cradle of diversity’, the centre of origin from which species appeared, radiated, and colonized other areas (Diels, 1908; Bews, 1927; Ingvar et al., 1968). In the other, tropical regions are considered a ‘museum of diversity’, where species suffered relatively fewer environmental disturbances over evolutionary time, allowing ancient lineages to be preserved for millennia (Simpson, 1980; Stebbins, 1974; Wallace, 1878). Although not mutually exclusive (McKenna and Farrell, 2006), the cradle vs. museum hypotheses primarily assume evolutionary scenarios in which diversity expands through time without limits (Hey, 1992). However, expanding diversity models may be limited in their ability to explain the entirety of the diversification phenomenon in the Neotropics; for example, expanding diversity models cannot explain the occurrence of ancient and species-poor lineages in the Neotropics (Condamine et al., 2015; Antonelli and Sanmartín, 2011b; Gibb et al., 2016) or the decline of diversity observed in the Neotropical fossil record (Hoorn et al., 1995; Jaramillo et al., 2006; Antoine et al., 2017). Although the concepts of cradle and museum have contributed to stimulate numerous macroevolutionary studies, a major interest is now focused on the evolutionary processes at play rather than the diversity patterns themselves (Vasconcelos and O’Meara, 2022). Four alternative evolutionary trajectories of diversity dynamics could be hypothesized to explain the accumulation of Neotropical diversity observed today (Figure 1):

Alternative hypotheses to explain current Neotropical diversity.

(a) Main species richness dynamics through time, and (b,c) the alternative evolutionary processes that could generate the corresponding patterns. (Sc. 1a) A gradual increase of species richness could result from constant speciation and extinction rates (1b), or through a comparable increase in speciation and extinction rates (1c). (Sc. 2a) An exponential increase in species numbers could be attained through constant extinction and increasing speciation (2b), or constant speciation and decreasing extinction rates (2c). (Sc. 3a) Saturated increase scenarios, with species accumulation rates slowing down towards the present, could result from constant extinction and decreasing speciation (3b), or through constant speciation and increasing extinction rates (3c). (Sc. 4a) Waxing and waning dynamics could result from constant extinction and decreasing speciation (4b), or constant speciation and increasing extinction (4c). Waxing and waning scenarios differ from saturated increases in that extinction exceeds speciation towards the present, such that diversification goes below 0. Scenarios (b–c) represent the simplest and most general models to explain species richness patterns in (a), but other combinations of speciation and extinction rates could potentially generate these patterns; for example, an exponential increase of species (2a) could also result from increasing speciation and punctual increases in extinction, or through increasing speciation and decreasing extinction.

Gradual expansions (Scenario 1)

This scenario proposes that species richness accumulated gradually through time in the Neotropics until the present, due, for example, to constant speciation and extinction rates. The gradual increase model received substantial support in the early and recent literature (Wallace, 1878; Couvreur et al., 2011; Derryberry et al., 2011; Santos et al., 2009; Schley et al., 2018; Harvey et al., 2020), and is generally associated with the long-term environmental stability and large extension of the tropical biome across the South American continent (Simpson, 1980; Stebbins, 1974).

Exponential expansions (Scenario 2)

An exponential increase in diversity model asserts that species richness accumulated faster towards the present. Such a pattern can result, for example, from constant extinction and increasing speciation rates, or constant speciation and decreasing extinction. Support for this model generally comes from studies suggesting that recent geological and climatic perturbations, mostly associated with the elevation of the Andes, promoted increases of diversification (Hoorn et al., 2010; Rull, 2011; Antonelli et al., 2018b). This diversity scenario is probably the most supported across Neotropical studies, although never quantified, with models of increasing speciation (Haffer, 1969; Richardson et al., 2001; Meseguer et al., 2020; Erkens et al., 2007; Hughes and Eastwood, 2006; Esquerré et al., 2019; Drummond et al., 2012; Lagomarsino et al., 2016; Pérez-Escobar et al., 2017; Musher et al., 2019; Olave et al., 2020) more often put forward than models of decreasing extinction (Antonelli and Sanmartín, 2011b).

Saturated or asymptotic expansions (Scenario 3)

A saturated diversity model postulates that species richness accumulated more slowly towards the present than in the past, reaching a diversity plateau. This can result from constant extinction and decreasing speciation, for example, such that speciation and extinction rates become equal towards the present. Diversification decreases could be due to ecological limits (Rabosky, 2009), damped increases (Cornell, 2013; Morlon et al., 2010), or abiotic fluctuations (Condamine et al., 2019a). Some studies support this model for the Neotropics, and they generally associate it with an early burst of diversification under favourable climatic conditions, followed by decelerations due to global cooling, and dispersal constraints (Santos et al., 2009; Phillimore and Price, 2008; Fine et al., 2014; Cadena, 2007; Weir, 2006).

Declines in diversity (Scenario 4)

Waxing and waning dynamics characterize clades that decline in diversity after periods of expansion. In a declining dynamic, diversification rates also decrease towards the present, but differ from saturated diversity in that extinction exceeds speciation, and diversity is lost. Waxing and waning dynamics may seem unlikely in a tropical context, but evidence for tropical diversity declines has been found at the global scale (Meseguer and Condamine, 2020; Quental and Marshall, 2013; Foote et al., 2007) and at the Neotropical scale in the fossil record (Hoorn et al., 1995; Jaramillo et al., 2006; Antoine et al., 2017; Archibald et al., 2010; Salas-Gismondi et al., 2015; Jansa et al., 2014; Carrillo et al., 2020). Fossil studies additionally suggest a link between decreases in Neotropical diversity and global temperature. For example, plant diversity inferred from fossil morphotypes reached its maximum levels during hyperthermal periods in the Eocene, and decreased sharply with subsequent cooling (Hoorn et al., 1995; Jaramillo et al., 2006; Wilf et al., 2005).

Despite an increasing number of evolutionary studies on Neotropical groups, today the prevalence of these alternative modes of species accumulation and diversification (Figure 1) at a continental scale has been difficult to tease apart empirically (Question 1). Yet, such an assessment would contribute to understand the origin and maintenance of Neotropical diversity. Illuminating the historical causes of Neotropical diversity further requires a closer look at the regional determinants of diversification. Are diversity trends (Sc. 1–4) related to specific environmental drivers (Question 2), geographic settings (Question 3), or taxonomic groups (Question 4) in the Neotropics?

Previous studies indicate that diversification rates might be structured geographically in the Neotropics (Harvey et al., 2020; Jetz et al., 2012; Quintero and Jetz, 2018; Rangel et al., 2018), with geography and climate being strong predictors of evolutionary rate variation (Quintero and Jetz, 2018; Rangel et al., 2018). For example, speciation may be high in regions subject to environmental perturbations, such as orogenic activity (Esquerré et al., 2019; Lagomarsino et al., 2016; Vasconcelos et al., 2020; Pouchon et al., 2018; Madriñán et al., 2013), and often not associated with current species richness (Harvey et al., 2020; Quintero and Jetz, 2018). Still, little is known on the geographic structure of long-term Neotropical diversification. Studies investigating spatial patterns of Neotropical diversification focus on long-term diversification dynamics of particular clades, for example, diversification trends of orchids across Neotropical regions (Pérez-Escobar et al., 2017), or cross-taxonomic patterns in shallow evolutionary time, that is, present-day speciation rates (Harvey et al., 2020; Quintero and Jetz, 2018; Smith et al., 2014). However, present-day speciation rates might not represent long-term diversification dynamics, especially when rates vary through time. Present diversification could be higher in one region than another without providing information on the underlying trend in diversification. Under time-variable rate scenarios, analysing diversity trends is crucial, but requires changing the focus from species to clades as units of the analyses. Unfortunately, there is still a lack of large-scale comparative data across wide phylogenetic and ecological contexts (Vasconcelos et al., 2020; Eiserhardt et al., 2017). Given the long history and vast heterogeneity of the Neotropics, general insights can only be provided if long-term patterns and drivers of diversification are shared among Neotropical lineages and areas.

This lack of knowledge may be also due to the challenge of differentiating between evolutionary scenarios based on birth-death models and phylogenies of extant species alone (Nee et al., 1994; Rabosky, 2010). Recent studies have raised concerns on difficulties in identifying parameter values when working with birth-death models under rate variation scenarios (Stadler, 2013; Burin et al., 2019), showing that speciation (birth, λ) and extinction (death, μ) rates sometimes cannot be inferred from molecular phylogenies (Louca and Pennell, 2020). This calls for (i) analysing ‘congruent’ models with potentially markedly different diversification dynamics but equal likelihood for any empirical tree (Louca and Pennell, 2020), or (ii) implementing a solid hypothesis-driven approach, in which a small number of alternative hypotheses about the underlying mechanism are compared against data (Morlon et al., 2022).

Based on an unparalleled comparative phylogenetic dataset containing 150 well-sampled species-level molecular phylogenies and 12,512 extant species, we evaluate the prevalence of macroevolutionary scenarios 1–4 (Figure 1) as general explanations for Neotropical diversification at a continental scale (Q1), their drivers (Q2), and their variation across biogeographic units (Q3) and taxonomies (Q4). To address Q3, we previously identify long-term evolutionary arenas of Neotropical diversification suitable for comparison. Depending on the taxonomic source (Raven et al., 2020; Meseguer et al., 2020), our dataset represents ~47–60% of all described Neotropical tetrapods, and ~5–7% of the known Neotropical plant diversity.

Results

Neotropical phylogenetic dataset

We constructed a dataset of 150 time-calibrated clades of Neotropical tetrapods and plants derived from densely sampled molecular phylogenies (Figure 2; Figure 2—source data 1; Meseguer, 2021). The dataset includes a total of 12,512 species, consisting of 6222 species of plants, including gymnosperms and angiosperms (66 clades, representing 5–7% of the described Neotropical seed plants); 922 mammal species (12 clades, 51–77% of the Neotropical mammals); 2216 bird species (32 clades, 47–59% of the Neotropical birds); 1148 squamate species (24 clades, 30–33% of the Neotropical squamates); and 2004 amphibian species (16 clades, 58–69% of the Neotropical amphibian diversity). Each clade in our dataset includes 7–789 species (mean = 83.4), with 53% of the phylogenies including more than 50% of the described taxonomic diversity (sampling fraction mean = 57%). Clade ages range from 0.5 to 88.5 million years (Myrs) (mean = 29.9; Figure 2—figure supplement 1). In this dataset, amphibian phylogenies are significantly larger than those of other clades (p<0.05) (Figure 2—figure supplement 1). Amphibian and squamate phylogenies are also significantly older (p<0). Groups also differ in sampling fraction: plant (p<0.01) and squamate (p<0) phylogenies are significantly less sampled than phylogenies of other groups. Our dataset triples the data presented in previous meta-analyses of the Neotropics in terms of number of species, for example, 214 clades and 4450 species in Antonelli et al., 2018c, and quadruples it in terms of sampling, with 20.8 species per tree in Antonelli et al., 2018c.

Figure 2 with 2 supplements see all
Time of origin for Neotropical tetrapods and plants.

Horizontal bars represent crown ages of 150 phylogenies analysed in this study. Shaded boxes represent the approximate duration of some geological events suggested to have fostered dispersal and diversification of Neotropical organisms. Inset histograms represent summary statistics for crown age (mean = 29.9 Myrs), sampling fraction (mean = 57%), and tree size (mean = 83.4 species/tree). Mean global temperature curve from Zachos et al., 2008. Abbreviations: K, Cretaceous; P, Paleocene; E, Eocene; O, Oligocene; M, Miocene; P, Pliocene (Pleistocene follows but is not shown); GAARlandia, Greater Antilles and Aves Ridge. Animal and plant silhouettes from PhyloPic (http://-phylopic.org/). Figure 2—source data 1 includes the dataset of plant, mammal, bird, squamate, and amphibian phylogenies and the original references for this data. Figure 2—figure supplement 1 represents summary statistics for crown age, sampling fraction, and tree size for each clade. Figure 2—figure supplement 2 includes box plots showing differences in sampling fraction, clade age, and number of species per tree for the different taxonomic groups considered in this study.

Figure 2—source data 1

Includes the dataset of plant, mammal, bird, squamate, and amphibian phylogenies and the original references for this data.

https://cdn.elifesciences.org/articles/74503/elife-74503-fig2-data1-v3.docx

Estimating the tempo and mode of Neotropical diversification

Diversification trends based on traditional diversification rates

To understand the tempo (Q1) and drivers of Neotropical diversification (Q2), we compared the fit of birth-death models applied to 150 phylogenies, including models where diversification rates are constant, vary through time, vary as a function of past global temperatures, or vary according to past Andean elevation (see Methods). When only models with constant diversification and time-varying rates were considered, constant models best fit 67% of the phylogenies (101 clades) (Supplementary file 1A). In the remaining 49 trees, we detected variation in diversification rates. Speciation decreased towards the present in 28 trees (57%), increased in 12, and remained constant (being extinction time-variable) in 9, although the proportions varied between lineages (Figure 3a). The proportion of clades that evolved at constant diversification decreased to 50.6% (76 clades) when the comparison included more complex environmental models (Figure 4; Supplementary file 1B; Meseguer, 2021). The proportion of time-variable models also increased to 74 trees.

Figure 3 with 1 supplement see all
Speciation trends in 150 phylogenies of Neotropical plants and tetrapods.

The histograms show the proportion of phylogenies for which constant vs. time-variable diversification models were the best fit, as derived from (a) canonical and (b) pulled diversification rates when comparing time-dependent models against constant models. In Figure 3a, the proportion of time-variable models is subdivided by the proportion of phylogenies in which speciation rates increase through time, decrease through time, or speciation remains constant (being extinction time-variable). In Figure 3b, speciation trends are derived from present-day pulled extinction rates μp(0): negative present-day pulled extinction rates values (μp(0)<0) indicate decreasing speciation trends through time (Louca and Pennell, 2020). Positive μp(0)>0 values are possible under both increasing and decreasing speciation rates, in which case speciation trends are designed as ‘unknown’. Figure 3—source data 1 provides the data to construct a and Figure 4a. Figure 3—source data 2 provides the data to construct Figure 3b. Figure 3—figure supplement 1 shows the proportion of phylogenies fitting different pulled diversification models for a reduced dataset including only trees with more than 20 species (N=99), or with a sampling fraction over 20% (N=137).

Figure 4 with 1 supplement see all
Diversity dynamics in 150 phylogenies of Neotropical plants and tetrapods.

The histograms show the proportion of phylogenies for which gradual increase (Sc. 1), exponential increase (Sc. 2), saturated increase (Sc. 3), and waxing and waning (Sc. 4) scenarios were the best fit, as derived from net diversification trends when comparing (a) time-dependent models against constant models and (b) environmental (temperature- and uplift-dependent models) against time-dependent and constant models. (c) Correspondence analysis showing the association between species richness dynamics (represented by blue arrows) and major taxonomic groups (red arrows). If the angle between two arrows is acute, then there is a strong association between the corresponding variables. Figure 4—source data 1 provides the data to construct Figure 4b and c. Source data to generate Figure 4a is provided as Figure 3—source data 1, file 2; Figure 4—figure supplement 1 shows the proportion of phylogenies best fitting different species richness dynamics for a reduced dataset including only trees with more than 20 species (N=99), or with a sampling fraction over 20% (N=137).

The empirical support for the main species richness dynamics from the 150 phylogenies was as follows: gradual expansions (Sc. 1, constant diversification) were detected in 101–76 phylogenies if environmental models were considered; exponential expansions (Sc. 2, increases in diversification) were detected in 20–30 clades; and saturated expansions and declining dynamics (Sc. 3 and 4, diversification decreases) were supported in 24–31 and 5–9 clades, respectively (Table 1 and Figure 4). Diversification trends remained similar when small (<20 species) or poorly sampled (<20% of the species sampled) phylogenies were excluded from the analyses (99 and 137 trees remaining, respectively), although the proportion of constant diversification models decreased in all cases (55–35%; Figure 3—figure supplement 1; Figure 4—figure supplement 1).

Table 1
Alternative species richness dynamics (Sc. 1–4) and the corresponding diversification processes (a–c) able to explain Neotropical diversity.

Species richness dynamics represent scenarios of expanding (Sc. 1–2), saturating (Sc. 3) and contracting (Sc. 4) diversity, in which speciation (λ) and/or extinction (μ) remain constant or vary through time. The number of phylogenies supporting each model is provided for all lineages pooled together, and for plants and tetrapods separately. Empirical support for each evolutionary model is based on canonical diversification rates (CDR), and pulled diversification rates (PDR), by comparing the constant model against different sets of time-variable models. For CDR, we provide as well the results (in italic) based on model comparisons including constant, time-variable, and paleoenvironmental-dependent (temperature and uplift) models.

Diversity dynamicsCDR all (plant/tetra)PDR all (plant/tetra)Diversification processModel parametersCDR all (plant/tetra)PDR all (plant/tetra)
Sc 1.
Gradual increase
101 (47/54)
76 (40/37)
95 (39/56)(a) Constant λ and μλ(t) = λ0,
μ(t) = μ0
101 (47/54)
77 (40/37)
95 (39/56)
(b) Equivalent increase in λ and μλ(t) = λ0eαt,
μ(t) = μ0eβt, λ0 = μ0,
α = β
0 (0/0)
0 (0/0)
(c) Both0 (0/0)
0 (0/0)
Sc 2. Exponential increase20 (17/3)
30 (19/11)
51 (25/26) *(a) Increasing λ, constant μλ(t) = λ0eαt,
α<0,
μ(t) = μ0
9 (7/2)
9 (8/1)
51 (25/26) *
(b) Constant λ, decreasing μλ(t) = λ0,
μ(t) = μ0eβt,
β>0
10 (10/0)
13 (11/2)
(c) Both1 (0/1)
8 (0/8)
Sc 3.
Saturated increase
24 (1/23)
31 (3/28)
(a) Decreasing λ, constant μλ(t) = λ0eαt,
α>0,
μ(t) = μ0
24 (1/23)
29 (3/27)
(b) Constant λ, increasing μλ(t) = λ0,
μ(t) = μ0eβt,
β<0
0 (0/0)
0 (0/0)
(c) Both0 (0/0)
1 (0/1)
Sc 4.
Waxing and waning
5 (1/4)
13 (5/8)
(a) Decreasing λ, constant μλ(t) = λ0eαt,
α>0,
μ(t) = μ0
1 (1/0)
(1/5)
(b) Constant λ, increasing μλ(t) = λ0,
μ(t) = μ0eβt,
β<0
1 (0/0)
(1/1)
(c) Both3 (0/4)
8 (3/2)
  1. *

    Pulled extinction rates (μp) can be useful for inferring speciation trends, for example, a negative present-day pulled extinction rate (μp(0)<0) is indicative that λ decreases through time. But the opposite is not necessarily true, that is, a positive present-day pulled extinction rate (μp(0)>0) does not necessarily indicate that λ increases through time (Louca and Pennell, 2020). Based on pulled extinction, we cannot infer either if diversification dropped below 0, and thus differentiate between the two scenarios in which λ decreases through time (3. damped increase and 4. waxing and waning dynamics). Similarly, based on pulled diversification rates, we cannot identify increases in speciation or time changes in extinction rates (scenarios 1b,c; 2a,b,c; 3b,c; 4b,c).

Rate variation was inferred from models that can capture the dependency of speciation and/or extinction rates over time (time-dependent models) or over an environmental variable (either temperature- or uplift-dependent models). Among them, temperature-dependent models explained diversification in 40 phylogenies (26.7%). Time-dependent models best fit 17 clades (11%). Uplift-dependent models explained another 11% (Figure 5, Supplementary file 1B). The relative support for time-, temperature- and uplift-dependent models remained similar regardless of whether we compared the fit of the best or second-best models (defined based on ΔAIC values; Figure 5—figure supplement 1), although overall support for constant-rate scenarios decreased in the latter.

Figure 5 with 2 supplements see all
Drivers of Neotropical diversification in 150 phylogenies of Neotropical plants and tetrapods.

The histograms report the proportion of (a) phylogenies whose diversification rates are best explained by a model with constant, time-dependent, temperature-dependent, or uplift-dependent diversification. The number of phylogenies (and species) per group is shown in parentheses. (b) The histograms report the number of phylogenies whose diversification rates are best explained by a model with constant, time-, temperature-, or uplift-dependent diversification according to different species richness scenarios (Exp = Exponential increase [Sc.2], Sat = Saturated increase [Sc.3], and Wax = Waxing and waning [Sc.4]), for plants, endotherm tetrapods, ectotherms, and all clades pooled together. (c) Correspondence analysis for the pooled dataset showing the association between species richness dynamics (represented by red arrows) and the environmental drivers (blue arrows). If the angle between two arrows is acute, then there is a strong association between the corresponding variables. Figure 5—source data 1 provides the data to construct this figure. Figure 5—figure supplement 1 shows the proportion of phylogenies best fitting different paleoenvironmental models based on the most supported and second most supported model. Results are also reported for a reduced dataset including only trees with more than 20 species (N=99), or with a sampling fraction over 20% (N=137). Figure 5—figure supplement 2 shows the comparison of diversification results based on different paleotemerature curves.

Figure 5—source data 1

Provides the data to construct this figure.

https://cdn.elifesciences.org/articles/74503/elife-74503-fig5-data1-v3.docx
Figure 5—source data 2

Shows diversification results for the most supported (lowest AIC value), and the second most supported diversification model.

https://cdn.elifesciences.org/articles/74503/elife-74503-fig5-data2-v3.docx

Results also remained stable regardless of the paleotemperature curve (Zachos et al., 2008; Hansen et al., 2013; Veizer and Prokoph, 2015) considered for the analyses (Figure 5—figure supplement 2). Diversification analyses based on the different paleotemperature curves produced almost identical results, in terms of model selection, parameter estimates, and main diversification trends. Therefore, we present and discuss the results based on the curve of Veizer and Prokoph, 2015, as this is the only curve spanning the full time range of all the Neotropical lineages included in our dataset (150 phylogenies).

Diversification trends based on pulled diversification rates

To gain further insights in Neotropical diversification (Q1), we explored congruent diversification models defined in terms of pulled diversification rates (PDR, rp) (Louca and Pennell, 2020; Louca et al., 2018). These analyses recovered consistent diversification trends with those found above: 63% of the phylogenies (95 clades) better fit constant pulled models (Figure 3b; Supplementary file 1C). Meanwhile in 37% of the phylogenies (55 clades) we found variation in PDR through time. Diversification trends remained similar when small (<20 species) or poorly sampled (<20% of the species sampled) phylogenies were excluded from the analyses (Figure 3—figure supplement 1). We also detected negative pulled present-day extinction rates μp(0) in most of the phylogenies (51 clades, 92%) in which PDR varied through time, suggesting that speciation was decreasing. Based on PDR, we could only detect constant diversification (Sc. 1) or decreases in speciation, and thus the combined support for Sc. 2, 3, and 4 (Table 1).

Neotropical bioregionalization

To examine the spatial variation of diversification dynamics within the Neotropics (Q3), we first had to identify geographic units of long-term Neotropical evolution suitable for comparison. We found that most clades in our study were distributed in most Neotropical WWF ecoregions (Figure 6—source data 1), suggesting that species presence-absence data might be of limited use for delimiting geographic units at the macroevolutionary scale of this study. In contrast, based on clades’ abundance patterns, we identified five clusters of regional assemblages that represent long-term clade endemism (Figure 6; Figure 6—figure supplement 1; Figure 6—source data 2): cluster 1 (including the Amazonia, Central Andes, Chocó, Guiana Shield, Mesoamerica, and Northern Andes), cluster 2 (Atlantic Forest, Caatinga, Cerrado, Chaco, and temperate South America), cluster 3 (Caribbean), cluster 4 (‘elsewhere’ region), and cluster 5 (Galapagos). An alternative clustering (Figure 6—figure supplement 2) separating Mesoamerica from cluster 1, and the Chaco and temperate South America from cluster 2, received lower support (Figure 6—figure supplement 1).

Figure 6 with 2 supplements see all
The geographical structure of long-term Neotropical diversification.

(a) Principal component analysis (PCA) representation of the five biogeographic clusters identified based on K-means clustering of 13 areas (WWF ecoregions) and 150 clades. (b) Resulting clusters (1–5) in geographic space. Colours correspond with the biogeographic clusters in (a). Thick lines delineate the original 13 ecoregions used in the analyses. (c) Box plot showing differences in crown age of the phylogenies distributed in each of the biogeographic clusters. (d) Variation in diversification and (e) pulled diversification rates (derived from the constant-rate model) across geographic clusters. (f) Number of phylogenies for which species richness scenarios Sc. 1–4 (Grad = Gradual increase [Sc.1], Exp = Exponential increase [Sc.2], Sat = Saturated increase [Sc.3], and Dec = Declining diversity [Sc.4]) were the best fit, across geographic clusters as derived from canonical diversification rates. (g) Number of phylogenies for which constant vs. declining speciation rates were the best fit, across geographic clusters as derived from pulled diversification rates. Figure 6—source data 1 provides the original data to conduct K-means clustering analyses, and generate Figure 6a; Figure 6—source data 2 provides the assignation of clades to biogeographic clusters; Figure 6—source data 3 provides the data to generate (c, d, f), and Figures 79; Figure 6—source data 4 provides the data to generate Figure 6; for example, Figure 6—figure supplement 1 shows the Elbow curve for K-means clustering results; Figure 6—figure supplement 2 shows biogeographic clustering and diversification results if seven clusters are considered.

Figure 6—source data 1

Provides the original data to conduct K-means clustering analyses, and generate Figure 6a.

https://cdn.elifesciences.org/articles/74503/elife-74503-fig6-data1-v3.docx
Figure 6—source data 2

Provides the assignation of clades to biogeographic clusters.

https://cdn.elifesciences.org/articles/74503/elife-74503-fig6-data2-v3.docx
Figure 6—source data 3

Provides the data to generate Figure 6c, d, f, and Figures 79.

https://cdn.elifesciences.org/articles/74503/elife-74503-fig6-data3-v3.docx
Figure 6—source data 4

Provides the data to generate Figure 6; for example.

https://cdn.elifesciences.org/articles/74503/elife-74503-fig6-data4-v3.docx

Variation of diversification dynamics across taxa, environmental drivers, and biogeographic units

We evaluated the prevalence of macroevolutionary scenarios 1–4 (Figure 1) across environmental drivers (Q2), biogeographic units (Q3) and taxonomies (Q4) (see Methods). Table 2 summarize all the results. We found that species richness dynamics were related to particular environmental drivers (p=0.003; Q2). Pairwise comparisons indicated that temperature-dependent models tended to best fit clades experiencing saturating (p=0.049) and declining (p=0.05) diversity dynamics. Meanwhile, uplift- and time-dependent models tended to best fit clades with exponentially increasing diversity (p=0.03) (Figure 5c).

Table 2
Summary p value results derived from the analysis of canonical diversification (r) and pulled diversification (rp) rates.

Significant differences in the proportion of clades experiencing different diversity trajectories (based on canonical diversification rates: gradual expansions, exponential expansions, saturation or declining diversity; based on pulled diversification rates: expanding vs. declining speciation) across biogeographic units, elevations, taxonomic groups, and environmental drivers as derived from Fisher’s exact tests. Significant differences in net diversification, pulled diversification, and speciation rates across biogeographic units, elevations and taxonomic groups derive from Kruskal-Wallis chi-squared analyses. Significant results are highlighted in bold.

Diversity trajectoriesDiversification ratesSpeciation rates
rrprrpr
Biogeographic units (5 clusters)0.4590.2520.1680.0830.248
Biogeographic units (7 clusters)0.5030.9470.1980.4240.277
Elevation0.5040.8390.6720.2770.034
Elevation (lowland-montane combined)0.0620.0620.3320.8690.031
Taxonomic groups0.0000.1260.0000.0000.000
Environmental drivers0.003

In contrast, there is no evidence to suggest that species richness dynamics are related to a given geographic location when considering the whole dataset (Figure 6c–f, Figure 7; Q3). Results of Fisher’s exact test show no significant differences in the proportion of clades experiencing gradual expansions, exponential expansions, saturation, or declining diversity across biogeographic units (p=0.45) or elevation ranges (p=0.062). We obtained similar results when the montane category was analysed separately (p=0.5, Figure 7). Diversity trajectories derived from the analysis of PDR produce the same results, with no differences in the proportion of clades experiencing constant (i.e., expanding diversity dynamics) or declining speciation trends across biogeographic units (p=0.25), or elevation ranges (p=0.062), even when the montane category was analysed separately (p=0.839). Estimates of net diversification rates (rather than diversity trajectories) derived from the constant diversification model did not differ across biogeographic units (χ2=5.05, p=0.17) or altitudinal ranges (χ2=2.20; p=0.332) either. Speciation rates did not differ between biogeographic units (χ2=4.1, p=0.25), but did vary across altitudinal ranges (χ2=6.9, p=0.03). Speciation rates were significantly higher across highland taxa (Figure 7). In addition, PDR did not differ across biogeographic units (χ2=6.7; p=0.083) or elevations (χ2=0.28; p=0.87).

Variation in diversification rates on 150 Neotropical phylogenies of plants and tetrapods across elevation ranges.

Diversification and speciation rates are derived from the constant-rate model (Supplementary file 1A). In the elevation code 1 the montane category has been analysed separately, while in the elevation code 2 lowland and montane categories have been pooled together (see text). Letters are used to denote statistically differences between groups, with groups showing significant differences in mean values denoted with different letters. Source data to generate this figure is provided as Figure 6—source data 3.

Finally, diversity trajectories (Sc. 1–4) differed across taxonomic groups (p<0.0001, Fisher’s exact test; Q4). Pairwise comparisons indicated that plants differed significantly from birds in the proportion of gradual (p<0.02), exponential (p<0.02), and saturated (p<0.0001) increase models after correcting for multiple comparisons. Birds also differed from amphibians in the proportion of saturated and exponential increases (p<0.02). Plants differed from squamates in the proportion of exponential (p<0.0006) and saturated (p<0.008) increases (Figure 4c). Net diversification rates were also significantly lower for Neotropical ectotherm tetrapods than for endotherms and plants (Kruskal-Wallis chi-squared: χ2=36.7, p<0.0001) (Figure 8). We also found statistically significant differences in speciation rates across groups (χ2=60.8, p<0.0001): plants showed higher speciation rates than endotherms, the latter, in turn, with higher speciation rates than ectotherms.

Diversification rates compared across plants and tetrapods (endotherms and ectotherms).

Diversification and speciation rates are derived from the constant-rate model . Letters are used to denote statistically differences between groups, with groups showing significant differences in mean values denoted with different letters. The y-axis was cut off at 1.0 to increase the visibility of the differences between groups. Upper values for plants are therefore not shown, but the quartiles and median are not affected. Units are in events per million years. Source data to generate this figure is provided as Figure 6—source data 3.

The number of species per phylogeny differed between model categories (phylogenetic ANOVA: F=10.9, p=0.002). Clades fitting gradual expansion models tended to have less species than clades fitting exponential (p=0.006) and declining (p=0.03) dynamics (Figure 9). Taxon sampling, however, did not differ significantly (F=4.5, p=0.53). Crown age differed between model categories, being on average younger for gradual scenarios than for exponential (p=0.03) and declining (p=0.03) dynamics.

Box plots showing differences in sampling fraction, clade age (i.e., crown age), and number of species per tree (i.e., tree size) for the phylogenies supporting gradual increase (Sc. 1) vs. exponential increase (Sc. 2) vs. saturated (Sc. 3) vs. declining diversity dynamics (Sc. 4).

Sampling fraction does not differ significantly between model categories, suggesting that there is no particular bias of sampling in our study. Meanwhile, there are differences in the tree size and crown age between model categories. Letters are used to denote statistically differences between groups, with groups showing significant differences in mean values denoted with different letters. Source data to generate this figure is provided as Figure 6—source data 3.

Finally, we found that no continuous (Kr = 0.06, p=0.6; Kλ=0.07, p=0.4; Krp = 0.07, p=0.6) or multi-categorical trait displays phylogenetic signal (Figure 10), suggesting that the distribution of trait values is not explained by the phylogeny itself.

Phylogenetic signal of different multi-categorical traits.

Inferred δ-values (in red) compared to the distribution of values when the trait is randomized along the phylogeny.

Discussion

Diversification dynamics in the Neotropics

Neotropical biodiversity has long been considered as being in expansion through time due to high rates of speciation and/or low rates of extinction (Stebbins, 1974; Harvey et al., 2020; Meseguer et al., 2020). Yet, to our knowledge, the generality of this trend in the Neotropics has not yet been evaluated or quantified. The higher support for the expanding diversity trend found here aligns with these ideas because most Neotropical clades (between 80% and 70%, if environmental models are considered) displayed expanding diversity dynamics through time (Figure 4; Table 1). Most of these clades experienced a gradual accumulation of lineages (Sc. 1; between 67% and 50%), and a lower proportion (14% and 16%) expanded exponentially (Sc. 2), thus diversity accumulation accelerated recently. Results based on PDR support these conclusions, with the largest proportion of clades expanding diversity (63%) due to gradual increases (Sc. 1; Figure 3).

Our results, however, also provide evidence that cradle/museum models are not sufficient to explain Neotropical diversity. Based on traditional diversification rates, 16–21% of the Neotropical clades, mostly tetrapods, underwent a decay in diversification, hence a slower accumulation of diversity towards the present (Sc. 3). While a pervasive pattern of slowdowns in speciation has been described at various geographic and taxonomic scales, for example, Morlon et al., 2010; Phillimore and Price, 2008; McPeek, 2008; Luzuriaga-Aveiga and Weir, 2019, Neotropical tetrapod diversity levels have only rarely been perceived as saturated (Santos et al., 2009; Harvey et al., 2020; Phillimore and Price, 2008; Weir, 2006). Furthermore, waxing-and-waning dynamics (Sc. 4) also characterize the evolution of 3–9% of the Neotropical diversity, consistent with paleontological studies (Hoorn et al., 1995; Jaramillo et al., 2006; Antoine et al., 2017). We found that the species richness of five plant and eight tetrapod clades declined towards the present (e.g., Sideroxylon [Sapotaceae], Guatteria [Annonaceae], caviomorph rodents, Thraupidae birds, or Lophyohylinae [Hylidae] frogs). This proportion might seem minor but is noteworthy when compared with the low support for this model found in the Neotropical literature, which could be explained by the difficulties in inferring negative diversification rates based on molecular phylogenies (Rabosky, 2010). Inferring diversity declines is challenging, and often requires accounting for among-clade rate heterogeneity (Morlon et al., 2011). As shown here, incorporating environmental evidence could also help identify this pattern, increasing support for this scenario relative to the comparisons without these models (Figure 4).

Clade age and size can partially explain the better fit of the constant diversification model, thus the gradually expanding trend (Sc. 1). However, these tree features cannot explain the relative support between time-varying increasing (Sc. 2) versus decreasing (Sc. 3, 4) scenarios (Figure 9). Constant diversification prevails among recently originated and species-poor clades in our study, as also shown in Condamine et al., 2019b, which could suggest that these clades had less time to experience changes in diversification. Alternatively, the power of birth-death models to detect rate variation decreases with the number of species in a phylogeny, as shown with different diversification approaches (Burin et al., 2019; Davis et al., 2013; Lewitus and Morlon, 2018), suggesting that tree size could hinder the finding of rate-variable patterns. The main patterns found in this study appear to be robust to sampling artefacts. The support for the expanding diversity scenario persisted (72–60% of clades) after excluding small trees from the analyses (<20 species; Figure 4—figure supplement 1). Then, the relative support for the exponentially expanding scenario (Sc. 2) increased at the expense of the gradually expanding scenario (Sc. 1), strengthening the generality of the expanding trend in the Neotropics.

Incomplete taxon sampling may flatten out lineages-through-time plots towards the present and artificially increase the detection of diversification slowdowns (Cusimano and Renner, 2010). If this artefact affected our results, we would expect to see that under-sampled phylogenies would tend to better fit saturated diversity models (Sc. 3). Instead, we found that sampling fraction did not differ between lineages fitting saturated versus expanding diversity models (Figure 9). Moreover, the proportion of clades fitting saturated models even increased (17–22%) after excluding poorly sampled phylogenies (<20% of the species sampled; Figure 4—figure supplement 1).

Support for decreasing diversification through time was larger when PDR were considered: 34% of the clades showed slowdowns in speciation (Figure 3). Based on PDR, however, we cannot infer if decay of speciation were accompanied by constant, declining or increasing extinction (Louca and Pennell, 2020), and thus determine the relative support for Sc. 2–4. If speciation slowdowns were accompanied by larger extinction decreases, it would be possible to recover expanding dynamics (Sc. 1, 2), but in most other cases, they would lead to declines in diversification (Sc. 3, 4). The limited interpretability of PDR prevents the extraction of further conclusions based on these rates (Morlon et al., 2022).

Still, our study illustrates the robustness of the diversification trend in the Neotropics to different modelling approaches. Despite parameter values varying substantially for some trees between the traditional and PDR methods (Supplementary file 1), a pattern also described in recent studies (Morlon et al., 2022), our analyses support a macroevolutionary scenario of expanding diversity for most Neotropical clades (Figure 4).

Taxon-specific patterns and drivers of Neotropical diversification

The variation in Neotropical diversification dynamics could be partially explained by the taxonomic affinity of the groups under study. Our study revealed contrasting evolutionary patterns for plants and tetrapods (Figure 4): diversity expansions (Sc. 1, 2) were more frequently detected in plants (~88%, 59 clades) than in tetrapods (~57%, 48 clades). In contrast, asymptotic increases (Sc. 3) were more frequent in tetrapods (33%, 28 clades) than in plants (4.5%, 3 clades; Tynanthus [Bignoniaceae], Chamaedoreae [Arecaceae], and Protieae [Burseraceae]). Net diversification rates were also higher in plants (Figure 8), in agreement with previous studies (Hernández-Hernández et al., 2021).

The study of PDR did not help to confirm or reject these conclusions. Rates from PDR models are significantly different between plants and animals (p≈0.00), in agreement with results based on traditional models (Table 2). Diversification trajectories derived from these rates are not different, with plants and animals exhibiting an equivalent fraction of phylogenies showing a decrease of speciation (Figure 3b). Since extinction dynamics cannot be derived from PDR models, we do not know if speciation slowdowns detected in plants were accompanied by larger extinction declines. Thus, we cannot rule out the scenario of expanding dynamics (Sc. 1, 2) for plants found based on traditional birth-death models.

Differences in the phylogenetic composition of the plant and tetrapod datasets do not explain this contrasted pattern. On average, plant phylogenies are not significantly younger or species-poorer than tetrapod phylogenies (Figure 2—figure supplement 2). Yet, the proportion of clades experiencing increasing dynamics is significantly higher for plants (Figure 4). Plant phylogenies are significantly less sampled than are tetrapod phylogenies, though, as explained above, incomplete taxon sampling tend to have the opposite effect over diversity curves: flattening out lineages-through-time plots towards the present, increasing the probability to detect saturated dynamics (Cusimano and Renner, 2010).

Alternatively, this contrasting evolutionary pattern may result from differential responses of plants and tetrapods to environmental changes (Figure 5). Global temperature change during the Cenozoic is found to be the main driver behind diversification slowdowns (Sc. 3) and declines (Sc. 4) of tetrapods, especially for endotherms (Figure 5). The positive correlation between diversification and past temperature in our temperature-dependent models indicates these groups diversified more during warm periods, such as the Eocene or the middle Miocene, and diversification decreased during cool periods. This result is in agreement with previous empirical studies (Condamine et al., 2019a; Moen and Morlon, 2014) and also with recent simulations showing a negative effect of climate cooling (and a positive effect of Andean orogeny; see below) on Neotropical tetrapod diversification (Hagen et al., 2021). According to the Metabolic Theory of Biodiversity, low temperatures can decrease enzymatic activity, generation times, and mutation rates (Gillooly et al., 2001), which may in turn affect diversification (Allen et al., 2006). Climate cooling may also decrease global productivity, resource availability, population sizes (Mayhew et al., 2012), or even species interactions (Chomicki et al., 2019). Only the New World monkeys (Platyrrhini) diversified more as temperature dropped. This could reflect the role of Quaternary events on primate speciation (Rull, 2011), and/or be an artefact of taxonomic over-splitting in this clade (Springer et al., 2012). In contrast, a few plant clades are influenced by temperature changes, with diversification increasing during the Neogene cooling (i.e., negative correlation between diversification and temperature; Figure 5). This opposite pattern suggests that Cenozoic environmental changes drove diversification slowdowns for some tetrapods, but stimulated plant diversification. Although Neotropical climate has been relatively stable through the Cenozoic in comparison to other regions (Ziegler et al., 2003; Morley, 2007), in the Neotropics, global cooling contributed to the expansion of several biomes, such as the alpine Paramos (Madriñán et al., 2013) and other open ecosystems (Cheng et al., 2013; Dick and Pennington, 2019), providing new opportunities for diversification. Higher mean speciation rates in plants than in tetrapods (Figure 8) could have provided plant lineages more opportunities for adaptation to changing environments (Hughes and Eastwood, 2006). Greater dispersal abilities in plants (Antonelli et al., 2018c; Sanmartín and Ronquist, 2004) may also explain this pattern.

Temperature changes emerge in our study as an important factor driving Neotropical diversification across macroevolutionary scales (Antonelli and Sanmartín, 2011a; Condamine et al., 2019a), but our results also reveal that this is not the only driver. A substantial proportion of diversification changes are attributed to Andean uplift and other factors (Figure 5). To a lesser extent, Neotropical diversification is explained by ecological limits on the number of species within a clade, which would imply that diversity is bounded by specific carrying capacities (Rabosky, 2009; Etienne et al., 2012). Among the tetrapod phylogenies supporting diversification slowdowns, time-dependent models explain 3% of them (four phylogenies; Figure 5, Supplementary file 1B), suggesting that ecological limits play a minor role in the Neotropics. Time-dependent models with decreasing speciation have been suggested to be a good approximation of diversity-dependent diversification, whereby speciation rates decline as species accumulate (Rabosky et al., 2014; Morlon, 2014). In fact, recent studies show that time- and diversity-dependent models are difficult to distinguish based on extant phylogenies (Pannetier et al., 2021). As discussed above, our results lend support to an alternative explanation for diversification slowdowns: the idea that tetrapods, for some periods, were less successful in keeping pace with a changing environment (Condamine et al., 2019a; Moen and Morlon, 2014).

The Andean orogeny mostly impacted tetrapod diversification (Hagen et al., 2021), especially ectotherms. Diversification of some lineages increased as the Andes rose, including Andean-centred lineages such as Liolaemidae lizards, but also others predominantly distributed outside the Andes, such as Leptodactylidae frogs. Sustained diversification in the context of Andean orogeny, both into and out of the Andean region, could be explained by increasing thermal and environmental gradients, from the equatorial areas to Patagonia or from west-east (Fouquet et al., 2014; Moen and Wiens, 2017). Other possible correlates include changes in elevational distributions of lineages (Kozak and Wiens, 2010; Hutter et al., 2017), or recurrent migrations (Santos et al., 2009; Esquerré et al., 2019).

In contrast to tetrapods, plant diversity expansions were primarily associated with temperature cooling and with time, where the latter represents a null hypothesis; the better fit of a time-dependent model, in comparison to environmental models, is generally indicative of factors not being investigated here (Morlon, 2014). Many of the plant lineages fitting time-dependent models represent textbook examples of ongoing radiations; for example, centropogonids (Lagomarsino et al., 2016), Lupinus (Drummond et al., 2012), or Inga Kursar et al., 2009, whose diversification has been associated with biotic drivers, such as the evolution of key adaptations or pollination syndromes. These factors are taxon-specific and were not evaluated in this study, where we focused on global phenomena. Similarly, we did not assess the role of the emergence of angiosperm-dominated rainforests in the evolution of tetrapods. Angiosperm-dominated forests were already established in the Neotropics by the Palaeocene (Carvalho et al., 2021), while the age of origin for most clades in our study postdates this period (Figure 2). In all cases, our results add support to the role of environmental and biotic factors as non-mutually exclusive drivers of macroevolutionary changes on Neotropical plants.

Neotropical bioregionalization at macroevolutionary scales

Understanding the spatial variation of Neotropical biodiversity dynamics is key to understanding the determinants of the exceptional diversity of the Neotropics. The first step towards this is the identification of evolutionary arenas of Neotropical diversification.

Conventional bioregionalizations schemes, such as biomes (Walter and Box, 1976), ecoregions (Olson et al., 2001), or other pre-defined biogeographic units (Antonelli et al., 2018c; Escalante et al., 2013; Morrone, 2014), could represent evolutionary arenas of diversification suitable for comparison. These bioregions have often been shown to be useful for categorizing actual species ranges. However, they are less appropriate for examining clade endemism at the macroevolutionary scale. The temporal origin of several bioregions postdates the origin of many of our clades (Figure 2). For instance, the Cerrado is inferred to have originated during the late Miocene (Simon et al., 2009), and the Chocó during the Pliocene-Pleistocene (Pérez-Escobar et al., 2019). In addition, most clades in our study appear distributed in most Neotropical ecoregions and could not be assigned to a single region (Figure 6—source data 1). The lack of a clear geographical structure for taxa of higher rank could be explained by the fact that conventional bioregionalizations generally represent categorizations based on data on the contemporary distribution of species without explicitly considering ancestral distributions or the relationships among species (Holt et al., 2013; Kreft and Jetz, 2010).

We propose an alternative bioregionalization scheme of the Neotropical region that accounts for long-term regional assemblages at macroevolutionary scales (Figure 6). We identify five biogeographic units that represent macroregions where different independent Neotropical radiations occurred over millions of years of biotic evolution. These regions are defined in terms of species richness patterns within clades (Figure 6—source data 1; Figure 6—source data 2), showing that species-rich clades in Amazonia also tend to be species-rich in the Andes, Chocó, Guiana Shield, and Mesoamerica (biogeographic cluster 1), without excluding that some species within these clades occur in other regions. Meanwhile, clades that are species-rich in the Atlantic Forest tend to be rich in the Caatinga, Cerrado, Chaco, and temperate South America (cluster 2). This regionalization roughly coincides with the Neotropical sub-regions proposed in previous studies (Morrone et al., 2022). The biogeographic cluster 1 corresponds with a broad ‘pan-Amazonian’ region that relied on the ancient Amazon Craton (Hoorn et al., 2010). Cluster 2 broadly groups different formations of the area known as the ‘Dry Diagonal’ (Prado and Gibbs, 1993; Luebert, 2021), which are geologically younger, dating from the Miocene (Pennington et al., 2006; Beerling and Osborne, 2006; Becerra, 2005). Although lineage crown ages do not differ between these regions (Figure 6). Clusters 1 and 2 include regions identified as transition zones in previous studies – Mesoamerica and temperate South America, respectively (Kreft and Jetz, 2013). Our analyses merged these regions with the core area with which it showed the greatest affinity, although other less supported classification schemes separate transition regions into individual clusters (Figure 6—figure supplement 1, Figure 6—figure supplement 2). Within each of these clusters, the contribution of in situ diversification is therefore more relevant than dispersion to explain their biotic assemblage. As such, these biogeographical clusters form distinctive units of Neotropical evolution and represent long-term clade endemism.

The geographical structure of Neotropical diversification

The variation in Neotropical diversification dynamics described in this study (Figure 4) could not be explained by geography. We did not find evidence to reject the null hypothesis of equal diversification, with similar diversity dynamics (Sc. 1–4) found across the biogeographic units of Neotropical evolution identified here (Figure 6, Table 2). We obtained the same result when Mesoamerica and temperate South America transition zones were analysed separately (Figure 6—figure supplement 2). In addition, we did not find differences in diversification dynamics between elevational ranges. These results were consistent whether we analysed net diversification rates or their derived diversity trends (Sc. 1–4). In the former, Neotropical lineages distributed in different elevations did differ in their speciation rates, as found in previous studies: speciation increased with altitude (Drummond et al., 2012; Weir, 2006; Quintero and Jetz, 2018; Vasconcelos et al., 2020; Rahbek et al., 2019). Elevated speciation rates might result from ecological opportunities on newly formed high-altitude environments, or those newly exposed after periods of cooling (Armijo et al., 2015; Blisniuk et al., 2005; Flantua et al., 2019). However, elevated speciation rates were also accompanied by elevated extinction in these habitats, hence net diversification remains comparable. The hypothesis of comparable diversification was also supported when comparing PDR (Figure 6). Geographic diversification may vary within taxonomic groups, though small sample sizes prevent us from drawing any firm conclusions on this.

The use of clades (rather than species) as evolutionary units in our biogeographic comparisons is original, and allowed to compare linage diversification trends through time (i.e., constant, expanding, declining) across regions, and not just present-day diversification rates, as in different comparable studies focused at the species level, for example, Harvey et al., 2020; Quintero and Jetz, 2018; Smith et al., 2014. Present-day diversification rates are structured geographically in the Neotropics (Harvey et al., 2020; Quintero and Jetz, 2018; Rangel et al., 2018), but our study shows that present diversification does not represent long-term evolutionary dynamics. The lack of a clear geographic structure of long-term diversification suggests that the evolutionary forces driving diversity in the Neotropics acted at a continental scale when evaluated over tens of millions of years. Evolutionary time and extinction could have eventually acted as levelling agents of diversification across the Neotropics over time.

These results also suggest that differences in species richness between the Neotropical bioregions defined here might not be attributable to long-term differences in diversification rates, nor to differences in diversification dynamics. Nor could time alone explain these differences, as we found no significant differences in the crown age of the phylogenies distributed in the different biogeographic clusters (Figure 6). Several studies have highlighted the role of dispersal in the configuration of modern Neotropical biotas (Carrillo et al., 2020; Smith et al., 2014; Bacon et al., 2015; Antonelli et al., 2015). By focusing exclusively on Neotropical radiations, we did not consider the role of dispersal into and out of the Neotropics (or within Neotropical regions) as an additional factor explaining Neotropical diversification. Future studies integrating biogeographic and diversification processes will be needed to provide a complete picture on the drivers of Neotropical diversification.

Limitations and perspectives

The results and conclusions presented here represent our best attempt to infer complex processes in deep geological times, and need to be interpreted in light of the general challenges in estimating diversification rates from phylogenies of extant species. Louca and Pennell, 2020, have reanimated this debate by showing that there is an infinite number of ‘congruent’ models that yield the same likelihood for any combinations of speciation and extinction rates. However, when speciation and extinction rates are defined as functions of time and constrained to follow specific functional forms, such as the exponential or a biologically motivated function (such as the environmental dependency tested here), speciation and extinction rates are identifiable (Morlon et al., 2022). The time-dependent models we applied have been shown to perform well in recovering speciation and extinction parameters, including negative net diversification (Morlon et al., 2011), detecting shifts of diversification (with regularization techniques as proposed in Morlon et al., 2022), and correctly identifying the diversification model and paleodiversity dynamic (Mazet et al., 2022). The same applies to environment-dependent models (Lewitus and Morlon, 2018).

According to previous simulations, it is unclear whether temperature-dependent models can be accurately distinguished when the effect of the environmental dependence on diversification is weak. Model selection tends to be sensitive when dependency values ranges between –0.1 and 0.1 (Lewitus and Morlon, 2018). In these cases, constant-rate models tend to overfit, which means that we are conservative when we conclude that temperature-dependent models are estimated as best fitting in our study. We therefore measured the impact this bias might have on our results, expecting that if the constant-rate model overfits, we would observe that the temperature-dependent model is more often ranked second in the selection procedure. Of the 76 clades with a constant-rate model as the best fit, our results indicate that 50% (38/76) have temperature-dependent models as the second best fit, 40% (30/76) have time-dependent models, and 10% (8/76) have Andean-dependent models (Figure 5—source data 2). This suggests that there is no clear bias against temperature-dependent models. Furthermore, when evaluating the dependency values of the 38 clades that are best fit by a constant-rate model and second best fit by a temperature-dependent model, we find that only 26% (10/38) have dependency values ranging between –0.1 and 0.1 for the temperature models. These 10 trees represent 6% of our dataset, suggesting that there is a low proportion of trees susceptible to suffer from this bias.

In our study, the relative support for time-, temperature- and uplift-dependent models remained stable to AIC variations (Figure 5—figure supplement 1). Model support also remained stable regardless of the paleotemperature curve considered for the analyses (Figure 5—figure supplement 2). Furthermore, the use of an hypothesis-driven framework has been suggested as a potential solution to alleviate the problem of non-identifiability of diversification parameters, by setting up explicit prior assumptions and delimiting the potential parameter space (Louca and Pennell, 2020; Morlon et al., 2022; Magee et al., 2020). Here, we do not evaluate every possible factor that could potentially explain Neotropical biodiversity, but only confront scenarios capturing well-established hypotheses on Neotropical diversification. We focus on the role of the Cenozoic change in Andean elevation and climatic oscillations because they have previously been pinpointed as essential for explaining Neotropical biodiversity (Hoorn et al., 2010; Rangel et al., 2018; Hagen et al., 2021). Thus, our main interest is to explore which of these factors likely explains the data compiled, although other factors could have played a role.

We have compiled as many Neotropical clades (and as many species per clade) as possible, resulting in a phylogenetic dataset representing, to our knowledge, one of the largest assembled to date. Yet, we are keenly aware that we still come up short, especially with the plant database. Our plant dataset (>6000 species, 66 clades) includes just a small fraction (~7% of the total species) of the vast diversity described in the region. As such, our results, which show contrasting diversification dynamics between plants and tetrapods, should be taken with caution. Future investigations would be necessary to confirm the generality of the expanding trend for plants. Basic knowledge of the real Neotropical diversity (and phylogenetic relationships) also remains incomplete, for example, Kier et al., 2005, and we anticipate the discovery of additional patterns by expanding the database.

Similarly, we did not manage to sample evenly across all regions. Our conclusions on the spatial patterns of diversification are derived from the study of a fraction of the Neotropical diversity, where tropical rainforest lineages from the broad ‘pan-Amazonian’ region are most abundant. Although sample size in our biogeographic comparisons is large (150 observations), some categories of these variables are poorly represented, which might limit the performance of some statistical tests. For instance, there are 97 phylogenies assigned to the biogeographic cluster 1, while only 10 in cluster 2. Note that there are other clades (39) containing species on poorly represented regions that fall in the ‘mixed’ category, as they share species with different areas. Our sampling, however, includes representatives from all the main regions in the Neotropics. Yet, we did not identify a common diversification trajectory or diversification rates, among the fewer clades distributed on poorly represented regions (e.g., southern South America clades experienced all gradual, exponential, and declining dynamics, as did the clades from other regions; Figure 6). It is also reasonable to assume that our sampling reflects a fair proportion of species per region, considering the extension of these regions in the Neotropics and the representativeness of our dataset; at least for tetrapods, it includes ~60% of all described species.

Although these limitations are likely to bias our study, we deem the representativeness of our dataset, and the diversification models compared here, as adequate to support the general patterns and conclusions inferred in this study. We hope that our study will provide interesting and testable perspectives for future investigations in the Neotropics and other regions.

Conclusions

This study represents a quantitative assessment of the prevailing macroevolutionary dynamics in the Neotropics, and their drivers, at continental and large temporal scales. Neotropical diversity has mostly expanded through time, but scenarios of saturated and declining diversity also account for a substantial proportion of Neotropical diversity. This variation in diversity trends is better explained by taxonomic rather than geographic factors, suggesting that the modern diversity observed in seed plants and tetrapods is partly a consequence of the contrasting diversification dynamics of these groups. Applying both traditional and pulled birth-death models to all phylogenies, we have shown a good consistency in the inferred models, which suggests that our study can provide meaningful estimates of diversification.

Whether the main pattern of diversity expansion over time can contribute to explain why the Neotropics have more species than other regions in the world remains to be evaluated based on comparative data from other regions (Antonelli et al., 2015; Couvreur, 2015). Such a comparison could reveal contrasted diversity trajectories in different continents and help to elucidate the association between current diversity levels and long-term diversity dynamics.

Methods

Data compilation

Neotropical clades, representing independent radiations in the Neotropics, were pulled from large-scale time-calibrated phylogenies of frogs and toads (Hutter et al., 2017), salamanders (Pyron et al., 2013; Pyron, 2014), lizards and snakes (Pyron and Burbrink, 2014), birds (Jetz et al., 2012) (including only species for which genetic data was available), mammals (Bininda-Emonds et al., 2007; Kuhn et al., 2011), and plants (Zanne et al., 2014). To identify independent Neotropical radiations, species in these large-scale phylogenies were coded as distributed in the Neotropics – delimited by the World Wide Fund for Nature WWF (Olson et al., 2001) – or elsewhere using the R package speciesgeocodeR 1.0–4 (Töpel et al., 2017), and their geographical ranges extracted from the Global Biodiversity Information Facility ‘GBIF’ (https://www.gbif.org/), the PanTHERIA database (https://omictools.com/pantheria-tool), BirdLife (http://www.birdlife.org), and eBird (http://ebird.org/content/ebird), all accessed in 2018, in a procedure similar to Meseguer et al., 2020. Next, we pruned the trees to extract the most inclusive clades that contained at least 80% Neotropical species, as previously defined. This procedure ensures that the diversification signal pertains to the Neotropics. In addition, phylogenies of particular lineages not represented in the global trees (or with improved taxon sampling) were obtained from published studies or reconstructed de novo in this study (for caviomorph rodents, including 199 species; Supplementary file 2). In the case of plants and mammals, most phylogenies were obtained from individual studies, given the low taxon sampling of the plant and mammal large-scale trees. However, whenever possible, we extracted phylogenies from a single dated tree rather than performing a meta-analysis of individual trees from different sources (Hoorn et al., 2010; Jansson et al., 2013), such that divergence times would be comparable. The resulting independent Neotropical radiations could represent clades of different taxonomic ranks. We did not perform any specific selection on tree size, crown age, or sampling fraction, but tested the effect of these factors on our results.

Estimating the tempo and mode of Neotropical diversification

Diversification trends based on traditional diversification rates

We compared a series of birth-death diversification models estimating speciation (λ) and extinction (μ) rates for each of the 150 phylogenies with the R package RPANDA 1.9 (Morlon et al., 2016) (Q1). To make these results comparable with those derived from PDR below, we followed a sequential approach by including models of increasing complexity. We first fitted a constant-rate birth-death model and compared it with a set of three models in which speciation and/or extinction vary according to time (Morlon et al., 2011): λ(t) and μ(t). For time-dependent models, we measured rate variation for speciation and extinction rates with the parameters α and β, respectively: α and β>0 reflect decreasing speciation and extinction towards the present, respectively, while α and β<0 indicate the opposite, increasing speciation and extinction towards the present.

We further compared constant and time-dependent models, described above, with a set of environment-dependent diversification models that quantify the effect of environmental variables on diversification (Q2) (Condamine et al., 2013). Environmental models extend time-dependent models to account for potential dependencies between diversification and measured environmental variables, for example, speciation and extinction rates can vary through time and both can be influenced by environmental variables. We focus here on mean global temperatures and Andean uplift. Climate change is probably one of the most important abiotic factors affecting biodiversity, of which global fluctuation in temperatures is the main component (Prokoph et al., 2008). In addition, the orogenesis of the Andes caused dramatic modifications in Neotropical landscapes and has become paradigmatic for explaining Neotropical biodiversity (Hoorn et al., 2010).

We fitted three environmental models in which speciation and/or extinction vary continuously with temperature changes (λ[T] and μ[T]), and three others with the elevation of the Andes (λ[A] and μ[A]). In this case, λ0 (μ0) is the expected speciation (extinction) rate under a temperature of 0°C (or a paleo-elevation of 0 m for the uplift models). We also analysed whether the speciation (α) and extinction (β) dependency were positive or negative. For temperature models, α(β)>0 reflects increasing speciation (extinction) with increasing temperatures, and conversely. For the uplift models, α(β)>0 reflects increasing speciation (extinction) with increasing Andean elevations, and conversely. We accounted for missing species for each clade in the form of sampling fraction (ρ) (Morlon et al., 2011) and assessed the strength of support of the models by computing Akaike information criterion (AICc), ∆AICc, and Akaike weights (AICω) to select the best fit model. We derived diversity dynamics (Sc. 1–4) based on the inferred diversification trends according to Figure 1.

For Andean paleo-elevations we retrieved a generalized model of the palaeo-elevation history of the tropical Andes, compiled from several studies (Lagomarsino et al., 2016 and references therein). The elevation of the Andes could have indirectly impacted the diversification of non-Andean groups. We thus applied uplift models to all clades in our study. Temperature variations during the Cenozoic were obtained from (i) global compilations of deep-sea oxygen benthic foraminifera (bf) isotope ratios (δ18Obf) (Zachos et al., 2008; Prokoph et al., 2008). This curve estimated by Prokoph et al., 2008, Veizer and Prokoph, 2015, and Zachos et al., 2008; Zachos et al., 2001 provides estimates for the last 540 Myrs, thus spanning the full time range over which Neotropical lineages diversified. However, recent investigations derived other paleotemperature curves for the Cenozoic (Hansen et al., 2013; Veizer and Prokoph, 2015; Cramer et al., 2011). To account for the uncertainty on global paleotemperatures on our results, we performed additional diversification analyses using other two different global curves; (ii) the temperature curve by Cramer et al., 2011; Cramer et al., 2009, which is similar to the more widely used previous curve but accounts for fluctuations in sea water (sw) δ18Osw through time and correct for ice volume. This curve provides temperature estimates for the last 62.4 Myrs; and (iii) the paleotemperature curve estimated by Hansen et al., 2013, for the last 65.6 Myrs, which accounts for ice volume and deep ocean temperature changes, and provides estimates of surface and deep-water temperature changes. These three different estimates mostly differ in the magnitude of the temperature changes but share the same overall trend (Figure 5—figure supplement 2). For this comparison, we only included groups overlapping the isotope record of the tree paleotemperature curves (<62.4 Myrs; resulting in 128 phylogenies).

Diversification trends based on pulled diversification rates

To gain further insights in Neotropical diversification (Q1), we explored congruent diversification models defined in terms of pulled diversification rates (PDR, rp), and pulled extinction rates (PER, μp) (Louca and Pennell, 2020; Louca et al., 2018). Two models are congruent if they have the same rp and the same product ρλ0, in which ρ is the sampling fraction and λ0 = λ(0). rp is equal to the net diversification rate (r = λ − μ) whenever λ is constant in time (/dτ=0) but differs from r when λ varies with time. The PER μp is equal to the extinction rate μ if λ is time-independent but differs from μ in most other cases. Pulled and canonical diversification parameters are thus not equivalent in most cases. Biological interpretation of pulled parameters is not obvious. However, some specific properties of PDR and PER allowed us to compare diversification dynamics estimated based on pulled and canonical diversification parameters. Specifically, changes in speciation and/or extinction rates usually lead to similarly strong changes in PDR, while constant PDR are strong indicators that both λ and μ were constant or varied only slowly over time (Louca and Pennell, 2020; Louca et al., 2018). PDR can also yield other valuable insights: if μp(0) is negative, this is evidence that speciation is currently decreasing over time (Louca and Pennell, 2020; Louca et al., 2018).

We estimated PDR values using the homogenous birth-death model on the R package castor 1.5.7 with the function fit_hbd_pdr_on_grid (Louca and Doebeli, 2018). We compared constant models (one time interval) with models in which PDR values are allowed to vary independently on a grid of three time intervals. We set up the age grid non-uniformly, for example, age points were placed closer together near the present (where information content is higher), and we selected the model that best explained the lineage-through-time of the Neotropical time trees based on AIC. To avoid non-global local optima, we performed 20 independent fitting trials starting from a random choice of model parameters. The fit_hbd_pdr_on_grid function additionally provided estimates of ρλo values. Knowing ρ, λ0 could be derived as follows: λ0 = λ0ρ/ρ. Similarly, pulled extinction rates for each time interval could be derived as follows: μp: = λ0 – rp. We limited the estimates to time periods with >10 species, using the oldest_age function in castor, to avoid points in the tree close to the root, where estimation uncertainty is generally higher.

Neotropical bioregionalization

We used a quantitative approach to identify geographic units of long-term Neotropical evolution. We divided the Neotropical region into 13 operational areas based on the WWF biome classification (Olson et al., 2001) and similar to other studies, for example, (Antonelli et al., 2018c; Hutter et al., 2017) – Amazonia, Atlantic Forest, Bahama-Antilles, Caatinga, Central Andes, Cerrado, Chaco, Chocó, Guiana Shield, Mesoamerica, the Northern Andes, temperate South America, and an ‘elsewhere’ region – and assessed the distribution in these areas of the 12,512 species included in our 150 phylogenies. Georeferenced records were downloaded for each species through GBIF using the R package rgbif 0.9.9 (Chamberlain et al., 2017). We removed points with precision below 100 km, entries with mismatched georeference and country, duplicates, points representing country capitals or centroids, using the R package CoordinateCleaner 1.0-7 (Zizka et al., 2019). Then, we created 13 georeferenced polygons delimiting each operational area using the WWF terrestrial ecoregions annotated shapefile in QGIS, and species were assigned to each polygon according to coordinate observations using the R package speciesgeocodeR 1.0-4. GBIF records can result in an overestimation of widespread ranges (Maldonado et al., 2015), so species distributions were manually inspected for completeness and accuracy with reference to databases (AmphibiaWeb 2018, Uetz et al. 2018, GBIF.org 2018, IUCN 2018). Based on the number of species belonging to each phylogenetic clade in the 13 ecoregions, we created a species abundance table (number of species per region per clade) that formed the basis for subsequent analyses.

The number of species distributed in each region within each clade were transformed using Hellinger transformations to account for differences in species richness between clades, and the Morisita-horn distance metric was selected to quantify pairwise dissimilarities of regional assemblages using the R package vegan 2.5-7 (Oksanen, 2013). We used K-means cluster analyses to form groups of similar regional assemblages. We determined the optimum number of groups by the elbow method. We use the function fviz_cluster in the R package factoextra 1.0.7 (Kassambara and Mundt, 2017) to visualize K-means clustering results using principal component analysis.

Variation of diversification dynamics across taxa, environmental drivers, and biogeographic units

We classified each clade in our study according to their main taxonomic group (plant [n=66], mammal [n=12], bird [n=32], squamate [n=24], amphibian [n=16]), environmental correlate (as estimated above: time [n=17], temperature [n=40] or uplift [n=17]), species richness dynamic based on canonical diversification rates (as estimated above: Sc. 1 [n=76], Sc. 2 [n=30], Sc. 3 [n=31], Sc. 4 [n=13]), and species richness dynamic based on PDR (constant speciation [n=83] and decreasing speciation [n=51]).

We also classified each clade into the biogeographic units identified above (see results): cluster 1 (including the Amazonia, Central Andes, Chocó, Guiana Shield, Mesoamerica, and Northern Andes, [n=97]), cluster 2 (Atlantic Forest, Caatinga, Cerrado, Chaco, and temperate South America, [n=10]), cluster 3 (Bahama-Antilles, [n=4]), cluster 4 (‘elsewhere’ region, [n=0]), or cluster 5 (Galapagos, [n=0]). Clades were assigned to a given cluster only if >60% of the species appeared in the cluster, otherwise clades were classified as ‘mixed’ (n=39; Figure 6—source data 2).

We additionally classified clades according to the main elevational range of their constituent species following literature descriptions rather than a purely quantitative approach as for the distribution above, because GBIF records in our dataset often came without associated altitude data (<30%): lowland [<1000 m; n=42] including lowland rainforest in Amazonia and the Chocó in western Colombia and Ecuador, as well as rainforest in the flanking lowland and pre-montane areas along the eastern side of the Andes; montane [1000–3500 m; n=8] including mid-elevation montane forests (e.g., cloud and elfin forests); highland [>3500 m; n=6] including alpine-altitude grasslands; mixed [n=94] includes lineages that show a mixed preference between lowland, montane and highland. Note that in our dataset, most clades fell into the mixed category, with montane species most often occurring within clades of lowland species, and rarely forming a clade of their own. To account for this pattern (and minimize the number of clades classified as ‘mixed’), we performed additional analyses pooling ‘lowland’ and ‘montane’ categories and considered a clade ‘mixed’ only if contained species in lowlands, montane and highlands (lowland-montane [n=124], highland [n=6], mixed [n=20]).

We assessed the phylogenetic signal of each multi-categorical trait (i.e., biogeographic units, elevation, diversity dynamics, and environmental drivers) using the δ-statistics (Borges et al., 2019) over a phylogeny including one tip for each of the 150 clades represented in this study. This tree was constructed using TimeTree (Kumar et al., 2017). High δ-value indicates strong phylogenetic signal. δ can be arbitrarily large, and thus significance was evaluated by comparing inferred δ-values to the distribution of values when the trait was randomised along the phylogeny. We evaluated the phylogenetic signal of continuous traits (i.e., diversification [r], speciation [λ], and pulled diversification [rp] rates) using Blomberg’s K (Blomberg et al., 2003) in the R package phytools 0.7–80 (Revell, 2012). Since time-varying diversification curves are hardly summarized in a single value, comparisons of net diversification values are based on estimates derived from the constant-rate model.

As no continuous or multi-categorical trait displays phylogenetic signal (see results), suggesting that the distribution of trait values is not explained by the phylogeny itself, statistical tests were conducted without applying phylogenetic corrections to account for the non-independence of data points. Fisher’s exact test was used in the analysis of contingency tables, performing pairwise-comparison with corrections for multiple testing (Benjamini and Hochberg, 1995), and Kruskal-Wallis tests for comparing means between groups.

We also tested the effect of clade age, size, and sampling fraction on the preferred species richness dynamic (Sc. 1–4) using a phylogenetic ANOVA in phytools with post hoc comparisons, checking if the residual error controlling for the main effects in the model and the tree were normally distributed. We applied phylogenetic corrections in this case because phylogenetic signal was detected for sampling fraction (Ksampling = 0.12, p=0.001) and crown age (Kage = 0.22, p=0.001), not for tree size (Ksize = 0.49, p=0.9).

Data availability

The chronogram dataset and the diversification results are archived in Dryad. All other data used or generated in this manuscript are presented in the manuscript, or its supplementary material.

The following data sets were generated

References

  1. Software
    1. Chamberlain S
    2. Ram K
    3. Barve V
    4. Mcglinn D
    5. Chamberlain MS
    (2017)
    Chamberlain
    Package ‘Rgbif.’.
    1. Condamine FL
    2. Rolland J
    3. Morlon H
    (2019b)
    Global climate cooling as an alternative explanation for diversification slowdowns
    Ecology Letters 22:1900–1912.
    1. Hansen J
    2. Sato M
    3. Russell G
    4. Kharecha P
    (2013) Climate sensitivity, sea level and atmospheric carbon dioxide
    Philosophical Transactions. Series A, Mathematical, Physical, and Engineering Sciences 371:20120294.
    https://doi.org/10.1098/rsta.2012.0294
  2. Book
    1. Mittermeier RA
    2. Turner WR
    3. Larsen FW
    4. Brooks TM
    5. Gascon C
    (2011)
    Global biodiversity conservation: the critical role of hotspots
    In: Zachos FE, Habel JC, editors. Biodiversity Hotspots. Springer-Verlag. pp. 3–22.
  3. Book
    1. Morley RJ
    (2007)
    Cretaceous and tertiary climate change and the past distribution of megathermal rainforests
    In: Bush MB, Flenley JR, editors. Tropical Rainforest Responses to Climatic Change. Springer. pp. 1–31.
    1. Nee S
    2. Holmes EC
    3. May RM
    4. Harvey PH
    (1994) Extinction rates can be estimated from molecular phylogenies
    Philosophical Transactions of the Royal Society of London. Series B, Biological Sciences 344:77–82.
    https://doi.org/10.1098/rstb.1994.0054
  4. Book
    1. Simpson GG
    (1980)
    Splendid Isolation: The Curious History of South American Mammals
    Yale University Press.

Decision letter

  1. David A Donoso
    Reviewing Editor; Escuela Politécnica Nacional, Ecuador
  2. Meredith C Schuman
    Senior Editor; University of Zurich, Switzerland
  3. Yaowu Xing
    Reviewer; Xishuangbanna Tropical Botanical Garden, China

Our editorial process produces two outputs: (i) public reviews designed to be posted alongside the preprint for the benefit of readers; (ii) feedback on the manuscript for the authors, including requests for revisions, shown below. We also include an acceptance summary that explains what the editors found interesting or important about the work.

Decision letter after peer review:

Thank you for submitting your article "The Origins and Drivers of Neotropical Diversity" for consideration by eLife. Your article has been reviewed by 3 peer reviewers, and the evaluation has been overseen by a Reviewing Editor and Meredith Schuman as the Senior Editor. The following individual involved in review of your submission has agreed to reveal their identity: Dr. Yaowu Xing (Reviewer #2).

The reviewers have discussed their reviews with one another, and the Reviewing Editor has drafted this to help you prepare a revised submission.

Essential revisions:

Please respond only to these essential revisions in your response letter, which have been determined through the consultative review process. In doing so you may refer to the comments in individual reviews that are related to the essential revisions.

1) We found potential inconsistencies between the title, your objectives, and your MandM. At first sight, reviewers could not see how the design of your study addresses why the Neotropics are so diverse relative to other regions, or why some regions of the Neotropics have more species than others. It is thus imperative that you temper your claims, and tone down your title.

2) Further, some reviewers expressed concern about the way divisions of the neotropics were done. Alternative splits of the Neotropical region (maybe in an appendix) could go a long way in easing concerns that traditional splits (Mesoamerica vs. Amazon vs. Andes) are not the best view of the diversity of the Neotropics.

3) Reviewers were puzzled by the lack of any confirmatory or accuracy-testing simulations on your methods. Without them, it is difficult at this point to evaluate if any strong bias exists in your results, or if your datasets have enough power to sustain your claims. It is essential that in a new version (and in your response letter) you address this issue. Does your methodology need simulations? Do currently available methods (e.g. Burin et al. 2019, Syst. Biol.; Louca and Pennell 2020, Nature) fill this gap?

Reviewer #2 (Recommendations for the authors):

In this study, the authors explored the evolution dynamics of Neotropical biodiversity by analyzing a very large data set, 150 phylogenies of seed plants and tetrapods. Furthermore, they compared diversification models with environment-dependent diversification models to seek potential drivers. Lastly, they evaluated the evolutionary scenarios across biogeographic regions and taxonomic groups. They found that most of the clades were supported by the expansion model and fewer were supported by saturation and declining models. The diversity dynamics do not differ across regions but differ substantially across taxa. The data set they compared is impressive and comprehensive, and the analysis is rigorous. The results broadened our understanding of the evolutionary history of the Neotropical biodiversity which is the richest in the world. It will attract broad interest to evolutionary biologists as well as the public who are interested in biodiversity.

The paper is well and clearly written in general.

1. My concern is about the sampling. It seems the authors only sampled the species that occurred in the Neotropics. It is OK for clades mostly distributed in this region. On the other hand, the sampling strategy overlooked the role of dispersal in generating biodiversity of Neotropics. However, many studies have already shown that dispersal (including long-distance dispersal) played important role in shaping Neotropical biodiversity. I suggest the authors discuss this somewhere in the Discussion part. Furthermore, I noticed that many tree sizes are small with less than 10 species. I do not know how reliable to estimate diversification rates for small-sized trees.

2. Though the authors quantified the effect of climate change, it seems the current resolution can not capture the role of Pleistocene climatic fluctuations which was proven to be important for particular biomes such as the alpine Paramos. Adding some discussions will strengthen the conclusions.

3. The first paragraph of the Conclusion part mainly illustrated the limitations of the study. I suggest the authors could put this part in the Discussions.

Reviewer #3 (Recommendations for the authors):

I was very impressed by the scope of the dataset, the thoroughness of the analyses, and the attention paid to caveats and limitations. I've provided a few comments below that are intended to strengthen this exciting manuscript.

One point I thought that might be of use to mention is that the focus on the recent past (very late Mesozoic-Cenozoic) means that South America's position relative to the present has varied little (ie. it was pretty much already in place by the time the study period begins), and this has implications of how much the climate has changed in these equatorial regions relative to more temperate regions (and the onset of the LDG during this time as global temps cooled).

L340-345: Were the smallest clades excluded also generally the youngest?

L356-357: Was this due to differences in clade size/age?

L364-375: Any idea if the discordance in endotherm diversification dynamics was due to New World monkeys vs. all other endotherms occupying a different range of latitudes? I'm just thinking about how lineages occupying more equatorial latitudes may have experienced a more limited drop in temperatures through time, relative to those at higher latitudes, and am wondering if this geographic separation (if it exists) could potentially explain the observed discordance?

L384-386: It might be useful to note here that while this may be the case, terrestrial tetrapods haven't gone completely extinct, as they've persisted since the Carboniferous.

L402-408: I like that attention is paid to potential biotic drivers of plant diversification; however, I'm also left wondering what role the evolution of angiosperm-dominated rainforests (which changed both the climate and vegetation structure) played any role in shaping the evolution of terrestrial tetrapods. I know it is beyond the scope of this paper to formally address it, but I think briefly mentioning the plausibility of this as a potential explanation for time-variable diversification dynamics would be beneficial.

L411-412: Any idea if this is due to the absence of other large clades restricted to the Neotropics, a lack of phylogenies for these, or because neotropical species are embedded in clades that have substantially radiated elsewhere (or some combination of these)?

L438: I think it's important to remember that these biomes, ecoregions and biogeographic units may be relatively recent.

L449-452: How can species richness patterns identify shared evolutionary histories? It feels as though it might be challenging to disentangle whether these richness patterns reflect in situ diversification in that biome or dispersal into that biome.

L498: uncomplete -> incomplete

L502: divers -> diverse

L524-531: Please clarify if any occurrence (including a single occurrence when 99% of occurrences are from outside the neotropics) was enough to code the taxon as occurring in the Neotropics, or if a higher threshold was set (xx% of occurrences from Neotropics). And the same request for assigning to taxa to the 13 WWF biomes.

L685-687: Was the 60% criterion applied above used here to determine whether clades were scored as mixed, lowland, montane, or highland?

L699-700: Doesn't this show that there is NOT phylogenetic signal (as it is written, it states that any trait contains phylogenetic signal, even though the K values are low and non-significant).

[Editors’ note: the authors submitted for reconsideration following the decision after peer review. What follows is the decision letter after the second round of review.]

Thank you for resubmitting the paper entitled "Diversification dynamics in the Neotropics through time, clades and biogeographic regions" for further consideration by eLife. Your revised article has been evaluated by a Senior Editor and a Reviewing Editor. We are sorry to say that we have decided that this submission will not be considered further for publication by eLife.

Both Reviewers and we (editors) valued the work done in the revision, but further comments raised by Reviewer 1 further weakened the validity of your approach. Specifically, the lack of a controlled methodology (by confirmatory or accuracy-testing simulations) made it difficult to evaluate if any strong bias exists in your results, or if your datasets have enough power to sustain your claims.

Reviewer #1 (Recommendations for the authors):

I previously reviewed this manuscript in December of 2021. I had several concerns. One was that the main method used to estimate diversification over time might be problematic. Another was that there was little comparison among different regions to understand patterns of diversity among regions in the Neotropics. I appreciate that the authors have made efforts to address these concerns, but I think that these problems remain somewhat problematic.

First, despite what the authors claim, I think the main approach used has not really been thoroughly tested with simulations. The authors claim that it has been, specifically by Morlon et al. (2011), Lewitus and Morlon (2018), and Condamine et al. (2019). The paper by Morlon et al. (2011) contains simulations, but (as far as I know) those simulations do not actually address the ability of the method to accurately choose among different models of diversification over time. The paper by Condamine et al. (2019) does not contain simulations at all. The paper by Lewitus and Morlon (2018) contains simulations, but primarily compares the ability of the method to distinguish time-dependent and temperature-dependent models (their Figure 3). But it is unclear if these two classes of models can be accurately distinguished from cases where the true model is a constant-rate model. Furthermore, looking at Figure 3 of Lewitus and Morlon (2018) it can be seen that this method often selects a constant-rate model when the true model is temperature dependent, if the effect of temperature dependence on diversification over time is not strong enough. The authors include in this list only papers co-authored by the developer of the method, and not other papers that found the method to be more problematic. It is also notable that the diversification dynamics are not significantly different between plants and animals when using PDR (Table 2). Therefore, not all of the main results are concordant between these two methods.

I found the comparison among biogeographic regions to be unsatisfying overall, because the regions are extremely coarse grained and each incorporates so many different regions. Furthermore, the comparison of 97 clades in one region (Mesoamerica+Amazonia+Andes and others) vs. only 10 clades in another (southern South America) also seems unsatisfying.

I appreciate that the authors describe their four scenarios explicitly in the Introduction, but I do not think that knowing the frequency of these four scenarios is "required to understand the origin and maintenance of Neotropical diversity" (line 112). Also, I do not see why the accumulation of species through "pulses" corresponds to the exponential expansion scenario. Should not constant speciation and low, constant extinction lead to an exponential increase in diversity over time? Even in Figure 1, I do not see the pulses in Scenario 2. A "pulse" implies a pattern that is discontinuous or episodic over time.

The finding that diversification patterns are different between plants and animals is interesting, but not entirely novel (see for example, Hernandez-Hernandez et al. 2021; Biological Reviews on the faster rates of diversification in plants vs. animals). It is also notable (again) that the diversification dynamics over time are not significantly different between plants and animals when using PDR, even if the rates are.

Finally, I am not sure how relevant these results are to future climate change, as is stated in the Abstract. It seems like this idea is not addressed in the paper itself.

Specific comments:

Lines 285 to 287: This needs to be rewritten since the results seem to contradict the conclusions.

Lines 340-342. The study cited focused on BiSSE-type models, not the approach used here.

Line 374. Change "plant phylogenies are significantly worst sampled than those of most other tetrapods" to "plant phylogenies are significantly less sampled than are tetrapod phylogenies"

Lines 415 to 416. How can there be microevolutionary studies of diversification?

Lines 538 to 540. The paper by Kreft and Jetz does not address diversification, only present-day species richness.

Lines 554 to 555. The "diverse range of diversification methods compared here" is actually only two diversification methods, by my count.

Line 563. What "substantial proportion of Neotropical diversity" is accounted for by these models?

Reviewer #2 (Recommendations for the authors):

The authors have addressed all of my comments. I am happy with the current version.

Reviewer #3 (Recommendations for the authors):

I thank the authors for their responses to my previous queries and approve of the changes made.

[Editors’ note: further revisions were suggested prior to acceptance, as described below.]

Thank you for choosing to send your work entitled "Diversification dynamics in the Neotropics through time, clades and biogeographic regions" for consideration at eLife. Your letter of appeal has been considered by a Senior Editor and a Reviewing Editor, and we are prepared to consider a revised submission with no guarantees of acceptance.

In this revised version, we stress that you must carefully address the following: (1) Please be more cautious about reporting the constraints of your chosen methods, (2) Please dilute the claim that you are resolving South American diversification, (3) Please address the several aspects of the most recent reviews which were not addressed in your appeal letter, including (but not limited to): your reliance on one specific method to calculate diversification rates, an apparent lack of simulations to support your chosen method, the coarse-grained resolution of the study and imbalance in clade numbers investigated in different regions, the request for a clearer and more critical projection of scenarios resulting from the proposed speciation patterns (do pulses necessarily lead to exponential expansion), and being careful to base conclusions about diversification patterns on dynamics and rates.

https://doi.org/10.7554/eLife.74503.sa1

Author response

Essential revisions:

Please respond only to these essential revisions in your response letter, which have been determined through the consultative review process. In doing so you may refer to the comments in individual reviews that are related to the essential revisions.

1) We found potential inconsistencies between the title, your objectives, and your MandM. At first sight, reviewers could not see how the design of your study addresses why the Neotropics are so diverse relative to other regions, or why some regions of the Neotropics have more species than others. It is thus imperative that you temper your claims, and tone down your title.

We thank you for your insightful comment and welcome these suggestions. We realize that the title we had initially chosen was so general that it might not reflect the content of the article. In our study we have four objectives: (1) to describe the main macroevolutionary dynamics that have prevailed in the Neotropics through time at a continental scale; (2) to assess their drivers; and (3) how they varied across biogeographic units and (4) taxonomic groups.

We agree with the editor and reviewers that we cannot directly answer why the Neotropics are more diverse than other regions of the world. To this end, we would need to compare our results (on the prevailing Neotropical dynamics) with comparative studies from other regions. We have changed the title of the study in the revised version as follows: “Diversification dynamics of plants and tetrapods in the Neotropics through time, clades and biogeographical regions”. We hope you will find this new title a better one reflecting the content of the article. In addition, to avoid any confusion, we have deleted the following sentence from the introduction: “But such an assessment is required to understand the origin of Neotropical diversity and why the Neotropics are more diverse than other regions in the world”.

Nevertheless, we would argue that our study could partially shed light on why some Neotropical regions have more species than others. Contrasted diversity levels across regions may be explained by three non-exclusive hypotheses: (1) species richness of a given region might be correlated to the amount of time available for speciation ("time-to-speciation" hypothesis; Stephens and Wiens, 2003); (2) some regions may just accumulate species at higher rates (diversification rate hypothesis); and (3) some regions may have more species because of higher rates of incoming dispersal (dispersal hypothesis). Assessing the relative support of these factors is not the main goal of our study, but our results allow us to evaluate the diversification rate hypothesis. We assess whether diversification rates vary across biogeographic units, and found similar diversification patterns and rates among regions. Thus, our results do not support the diversification rate hypothesis, and suggest that other factors might be at play. We have included this additional rationale in our Discussion: “Our results therefore suggest that differences in species richness between the Neotropical bioregions identified here might not be attributable to differences in diversification rates, and that other factors such as time (127), or asymmetric dispersal (4) could explain this pattern”.

2) Further, some reviewers expressed concern about the way divisions of the neotropics were done. Alternative splits of the Neotropical region (maybe in an appendix) could go a long way in easing concerns that traditional splits (Mesoamerica vs. Amazon vs. Andes) are not the best view of the diversity of the Neotropics.

Thank you for this comment. We have to admit that the division of the Neotropics into areas reflecting clade (and not species) endemicity has been the most complicated part of the study. There is a lot of discussion in the literature on the bioregionalization of species ranges and no consensus (Antonelli et al., 2018 – PNAS). When it comes to the bioregionalization of clade ranges, there is very few information and harder decisions to make. We started coding each phylogeny following classical bio-regionalization schemes (e.g., biomes, ecoregions), but quickly realized that they were not useful at the scale of this study, since most clades have species distributed in most Neotropical ecoregions. Thus, for previous versions of this manuscript, we decided to categorize clades within the Neotropics based on the three dominant patterns described by Gentry (1982 – Annals Miss. Bot. Gar.) that roughly characterize the main distribution of species richness, as well as the colonization history of Neotropical lineages: Andean-centered = lineages with species richness mostly distributed in the Andes, Chocó and Central America, and poorly represented in Amazonia, Atlantic Forest, Caatinga and Cerrado; Amazonian-centered = species richness mostly distributed in the Amazonia and Brazilian shield and poorly represented in Central America; and other = distribution that does not match the two previous categories (e.g. Southern cone of South America). We then compared the diversification dynamics between these regions and found similar results to those presented in this manuscript: the prevailing diversification dynamics and rates do not change between Gentry’s regions (see Author response image 1).

Author response image 1
Variation in diversification rates on 150 Neotropical phylogenies of plants and tetrapods across geographic ranges.

Diversification rates are derived from the constant-rate model.

However, as could be noticed, this classification is quite subjective: “… lineages with species richness MOSTLY or POORLY distributed in this or the other area” are rather subjective terms, which could lead other researchers to apply different criteria to categorize the same phylogenetic tree. Other traditional splits, as suggested here into Mesoamerica vs. Amazon vs. Andes, are even more subjective. How should we classify clades with species distributed in the Chocó, or in the Cerrado, for example, under this scheme?

Therefore, in this study, we decided to reduce all subjective divisions and use a quantitative approach to identify macroevolutionary Neotropical bioregions. We think that this choice is one of the greatest strengths of our study, as it makes our results repeatable and more objective than previously. In addition, by using this quantitative approach, we have revealed new patterns not identified before: for example, we have found that the Andes, Amazonia, and Chocó have acted as a single geographic unit of biotic evolution at the clade level for tens of millions of years. The existence of regional arenas of evolution in the Neotropics at the clade level and at large time scales is undoubtedly a key finding of this study. If the editor agrees, we would prefer not to include alternative and less objective divisions of the Neotropics in the study, which, as we demonstrate with our study, do not reflect the endemism patterns of Neotropical clades, and might thus lead to confusion.

3) Reviewers were puzzled by the lack of any confirmatory or accuracy-testing simulations on your methods. Without them, it is difficult at this point to evaluate if any strong bias exists in your results, or if your datasets have enough power to sustain your claims. It is essential that in a new version (and in your response letter) you address this issue. Does your methodology need simulations? Do currently available methods (e.g. Burin et al. 2019, Syst. Biol.; Louca and Pennell 2020, Nature) fill this gap?

This is a legitimate comment, and we understand the skepticism on a study that relies on macroevolutionary models of questionable robustness (e.g. Kubo and Iwasa 1995 – Evolution; Rabosky and Lovette 2008 – Evolution; Crisp and Cook 2009 – Evolution; Quental and Marshall 2010 – TREE; Burin et al. 2019 – Syst. Biol.; Louca and Pennell 2020 – Nature; Pannetier et al. 2021 – Evolution).

The methodology used here has been thoroughly tested with both simulations (e.g. Morlon et al. 2011 – PNAS; Lewitus and Morlon 2018 – Syst. Biol.; Condamine et al. 2019 – Ecol. Lett.) and empirical cases (e.g. Lewitus et al. 2018 – Nat. Ecol. Evol.; Condamine et al. 2019 – Ecol. Lett.). We cannot deny that such a methodology is fully free from issues, which affect all birth-death models, and brings the question: are we able to reliably infer the diversification model and identify parameter values of this model (Louca and Pennell 2020 – Nature)? These concerns are not likely to be resolved in the short term. Although many studies are making progress in understanding the behavior of diversification rate functions, showing, for example, that equally likely diversification functions (i.e. the congruent parameter space of Louca and Pennell 2020 – Nature) can share common features, with diversification rate patterns being robust despite non-identifiability (Höhna et al., 2022 – bioRxiv; Morlon et al., 2022 – TREE).

Being aware of these concerns, we also relied on the recently developed Pulled Diversification Rates method (Louca and Pennell 2020 – Nature; Louca et al., 2018 – PNAS) that is supposed to correct for the identifiability issue raised by recent studies. Hence, applying both traditional and pulled birth-death models to all phylogenies, we have shown a good consistency in the inferred models, which suggests that our study can provide meaningful estimates of diversification. Our empirical study is also one of the first to perform such a large-scale methodological comparison in diversification analyses (pulled vs. traditional birth-death models) while addressing a key question in evolutionary biology. We have now emphasized this point in the conclusions of our study: “To the extent possible, these results are based on traditional diversification rates, and on the recently developed Pulled Diversification Rates method that is supposed to correct for the identifiability issue raised by recent studies associated with traditional diversification rates (71). Hence, applying both traditional and pulled birth-death models to all phylogenies, we have shown a good consistency in the inferred models, which suggests that our study can provide meaningful estimates of diversification”.

Reviewer #2 (Recommendations for the authors):

In this study, the authors explored the evolution dynamics of Neotropical biodiversity by analyzing a very large data set, 150 phylogenies of seed plants and tetrapods. Furthermore, they compared diversification models with environment-dependent diversification models to seek potential drivers. Lastly, they evaluated the evolutionary scenarios across biogeographic regions and taxonomic groups. They found that most of the clades were supported by the expansion model and fewer were supported by saturation and declining models. The diversity dynamics do not differ across regions but differ substantially across taxa. The data set they compared is impressive and comprehensive, and the analysis is rigorous. The results broadened our understanding of the evolutionary history of the Neotropical biodiversity which is the richest in the world. It will attract broad interest to evolutionary biologists as well as the public who are interested in biodiversity.

The paper is well and clearly written in general.

1. My concern is about the sampling. It seems the authors only sampled the species that occurred in the Neotropics. It is OK for clades mostly distributed in this region. On the other hand, the sampling strategy overlooked the role of dispersal in generating biodiversity of Neotropics. However, many studies have already shown that dispersal (including long-distance dispersal) played important role in shaping Neotropical biodiversity. I suggest the authors discuss this somewhere in the Discussion part. Furthermore, I noticed that many tree sizes are small with less than 10 species. I do not know how reliable to estimate diversification rates for small-sized trees.

Thank you for pointing this out. Our sampling indeed makes a focus on clades endemic to the Neotropics, i.e. independent Neotropical radiations, which by definition includes in-situ diversification and limits the role of dispersal into the Neotropics. This procedure ensures that the diversification signal we analyze pertains to the Neotropics and not to other areas. We do not neglect dispersal as a driver, but for that we would need to rely on a biogeographical approach that is beyond the scope of this study. In the revised manuscript, we have acknowledged that dispersal into and out of the Neotropics is an additional factor that we did not take into account in this study. We have added the following sentence in the discussion: “Furthermore, by focusing exclusively on Neotropical radiations, we did not consider the role of dispersal into and out of the Neotropics (or within Neotropical regions) as an additional factor explaining Neotropical diversity. In fact, several studies have suggested an important role of dispersal in the configuration of modern Neotropical biotas (Smith et al., 2014; Carrillo et al., 2020; Bacon et al., 2015). Future studies integrating biogeographic and diversification processes will be needed to bridge this gap, and to provide a more complete picture on the drivers of Neotropical diversity.” Please see also above Essential revision 1.

Estimating rates of diversification for small-sized trees is undeniably difficult and comes with more uncertainties. However, we wanted to include these small phylogenies because they also represent the Neotropical biodiversity. We did not want to introduce a bias in our meta-analysis toward large-sized trees that would only show variable-rate models and would hide a substantial part of the biodiversity associated with this diversification pattern. We also show that constant-rate models can explain the diversification in the Neotropics. There is a more balanced diversification pattern, and we think it is fairer to present both large and small phylogenies. Nevertheless, we agree that small trees are problematic. We have evaluated the effect of including small trees in our sample. As shown in our Figure 9, clades fitting gradual expansion models tend to have less species than clades fitting exponential (p=0.006) and declining (p=0.03) dynamics, suggesting that small trees could introduce a bias in our results. We therefore repeated all the analyses excluding small trees (<20% of the species sampled) from our analyses, and found that the largest support for the expanding diversity trend persisted (72–60% of clades).

We discuss the difficulty to estimate rates of diversification from small-sized phylogenies and the possible bias introduced by the small phylogenies in our study in the Discussion section: “It has been suggested that the power of birth–death models to detect rate variation can decrease with the number of species in a phylogeny (Davis et al., 2013 – BMC Evol. Biol.), suggesting that tree size could hinder the finding of rate-variable patterns. Still, the patterns found in this study appear to be robust to sampling artifacts. We have repeated all the analyses excluding small (<20 species) and poorly sampled (<20% of the species sampled) phylogenies from our analyses, and found that the largest support for the expanding diversity trend persisted (72–60% of clades). Then, the relative support for the exponentially expanding scenario (Sc. 2) increased at the expense of the gradually expanding scenario (Sc. 1), strengthening the conclusion of the generality of the expanding trend in the Neotropics”. We hope you will find this result sufficient to demonstrate that the consideration of small trees in our analyses does not drive our main conclusions.

2. Though the authors quantified the effect of climate change, it seems the current resolution can not capture the role of Pleistocene climatic fluctuations which was proven to be important for particular biomes such as the alpine Paramos. Adding some discussions will strengthen the conclusions.

Thank you for this comment. We agree that Pleistocene climatic changes have had notable effects on speciation and extinction around the world. This has been largely studied and documented, and our study aims at extending the time frame without ignoring these findings. First of all, note that the temperature data we used includes the Pleistocene climatic fluctuations (e.g. Zachos et al. 2001 – Science; Zachos et al. 2008 – Nature), but it is true that smoothing the temperature curve tends to erase the important features of the Pleistocene fluctuations. In addition, the birth-death models used here cannot fit well these fine-scale variations because there are a lot of temperature ups and downs over a short time interval, while the number of speciation events in the same time frame is often low. We think it is probably illusory to believe that macroevolutionary models can provide reliable inferences for microevolutionary processes. Our study aims at providing estimates on a longer time-scale, which is often lacking compared to microevolutionary studies that have yielded numerous examples of the effect of Pleistocene climatic changes on biodiversity. That being said, we have revised the manuscript to discuss the role of recent climatic changes. We have included the following sentences: “The role of recent climate change over Neotropical diversity has been widely documented, especially concerning the impact of Pleistocene climatic fluctuations on tropical forests (Haffer 1968; Rull 2011), or in the formation of species-rich biomes such as the Paramo (Madriñan et al., 2019). Recent climatic variations are probably not sufficiently captured in our study. Our diversification models cannot fit well the fine-scale variations of the Pleistocene fluctuations because there were a lot of temperature ups and downs over a short time interval, while the number of speciation events in the same time frame can be low. In addition, the temperature curve is smoothed when incorporated into the model, which probably erases some of the important features of these fluctuations. Our study aims at providing estimates on a longer time-scale, which is often lacking in microevolutionary studies. Although Neotropical climate has been relatively stable through the Cenozoic in comparison to other regions (94, 95) – in the Neotropics global cooling contributed to the expansion of several biomes, such as the alpine Paramos (62) or open ecosystems (96, 97) – temperature changes emerge in our study as an important factor driving Neotropical diversity across macroevolutionary scales. These results are in line with previous studies suggesting that temperature is a key driver of biodiversity change at different evolutionary scales (7, 43). However, our results also reveal that climate change is not the only factor shaping Neotropical diversity at macroevolutionary scales.”

3. The first paragraph of the Conclusion part mainly illustrated the limitations of the study. I suggest the authors could put this part in the Discussions.

Thank you for this suggestion. We agree and have revised the main text accordingly.

Reviewer #3 (Recommendations for the authors):

I was very impressed by the scope of the dataset, the thoroughness of the analyses, and the attention paid to caveats and limitations. I've provided a few comments below that are intended to strengthen this exciting manuscript.

Thank you for your review, the positive input and all the comments.

One point I thought that might be of use to mention is that the focus on the recent past (very late Mesozoic-Cenozoic) means that South America's position relative to the present has varied little (ie. it was pretty much already in place by the time the study period begins), and this has implications of how much the climate has changed in these equatorial regions relative to more temperate regions (and the onset of the LDG during this time as global temps cooled).

This is an important point to make. Indeed, the timeframe of the study is mostly focused on a period when South America was approximately at the same place and latitudes as today. It broke up from Africa around 110 million years ago (e.g. Seton et al. 2012 – Earth-Sci. Rev.; Müller et al. 2016 – Annu. Rev. Earth Planet. Sci.), and slowly moved to reach its current position. However, we don't think this factor explains the pattern found here, as the latitudinal position of other continental landmasses also varied little during the Cenozoic. Except for Africa, which collided with the Arabian plate during the Miocene, North America and Eurasia were also pretty much in the same latitudinal position during the entire Cenozoic (Sanmartin et al. 2001 – Biol. J. Linn. Soc.; Seton et al. 2012 – Earth-Sci. Rev.; Müller et al. 2016 – Annu. Rev. Earth Planet. Sci.). It is true that global Cenozoic cooling was much less intensively felt in South America than in other regions, and the region remained tropical since the beginning of the Cenozoic, although global cooling contributed to the expansion of more open ecosystems (Cheng et al. 2013 – Nat. Com.; Dick and Pennington 2019 – Annu. Rev. Ecol. Evol. Syst.). In the revised version of the main text, we have mentioned this aspect pertaining to South America, its tropicality, and the effect of climate change at tropical latitudes:

“The role of recent climate change over Neotropical diversity has been widely documented, especially concerning the impact of Pleistocene climatic fluctuations (Haffer 1968; Rull 2011) […] Our study aims at providing estimates on a longer time-scale, which is often lacking in microevolutionary studies. Although Neotropical climate has been relatively stable through the Cenozoic in comparison to other regions (94, 95) – in the Neotropics global cooling contributed to the expansion of several biomes, such as the alpine Paramos (62) or open ecosystems (96, 97) – temperature changes emerge in our study as an important factor driving Neotropical diversity across macroevolutionary scales. These results are in line with previous studies suggesting that temperature is a key driver of biodiversity change at different evolutionary scales (7, 43)”.

L340-345: Were the smallest clades excluded also generally the youngest?

We did not observe any relationship between clade age and number of species. Models assuming constant speciation and extinction rates through time best fit half of the phylogenies in our study, often when clades are species-poor. Phylogenies best fitting a constant model are also significantly younger. However, young clades do not tend to be species-poor (Pearson's r = 0.35; see Author response image 2). We did not find a correlation between the other variables either; crown age – sampling fraction (r = 0.15), sampling – number of tips (r = 0.11).

Author response image 2
Scatterplots showing the relationship between (log transformed) sampling fraction, clade age (crown age in million years ago), and number of tips for the 150 phylogenies examined in this study.

The plots shows no association between the variables.

L356-357: Was this due to differences in clade size/age?

This is a very good point. We agree this comparison is relevant to support our conclusions, but it was missing from our results. We have now compared tree size, crown age and sampling fraction across taxonomic groups, and found that the higher proportion of increasing dynamics, characteristic of plants, cannot be explained by significant differences in these factors. As can be seen on the figure below (new Figure-2—figure supplement 2 on the manuscript), tree size does not differ among plants, mammals, birds and squamates. Crown age does not differ among plants, mammals and birds. Groups do differ on sampling fraction: plant (p < 0.01) and squamate (p < 0) phylogenies are significantly worst sampled than the phylogenies of other groups. Yet plants show a higher frequency of increasing dynamics than squamates, and other tetrapods (Figure 4). Incomplete taxon sampling has the effect of flattening out lineages-through-time plots towards the present, and thus artificially increasing the detection of diversification slowdowns rather than diversification increases (Cusimano and Renner 2010 – Syst. Biol.).

We have included this important piece of information in the results “In our dataset, amphibian phylogenies are significantly larger than those of other clades (p < 0.05) (Figure 2 —figure supplement 2). Amphibian and squamate phylogenies are also significantly older (p < 0). Groups also differ in sampling fraction: plant (p < 0.01) and squamate (p < 0) phylogenies are significantly worst sampled than phylogenies of other groups.”; and in the Discussion section: “Differences in the phylogenetic composition of the plant and tetrapod datasets do not explain this contrasted pattern. On average, plant phylogenies are not significantly younger or species-poorer than tetrapod phylogenies (Figure 2 —figure supplement 2). Yet, the proportion of clades experiencing increasing dynamics is significantly higher for plants (Figure 4). Plant phylogenies are significantly worst sampled than those of most other tetrapods, though, as explained above, incomplete taxon sampling has the opposite effect: flattening out lineages-through-time plots towards the present (83).”

L364-375: Any idea if the discordance in endotherm diversification dynamics was due to New World monkeys vs. all other endotherms occupying a different range of latitudes? I'm just thinking about how lineages occupying more equatorial latitudes may have experienced a more limited drop in temperatures through time, relative to those at higher latitudes, and am wondering if this geographic separation (if it exists) could potentially explain the observed discordance?

Very interesting point. A priori, our data do not suggest a contrast in the latitudinal distribution of New World monkeys vs. other endotherms to explain this difference. Most species of New World monkeys appear distributed in tropical latitudes, and this clade was assigned to our biogeographic cluster 1 (the “Pan-Amazonia” macroregion). We quantified that most species of monkeys occur in the Amazonia (61 spp.), Northern Andes (25), Guiana Shield (17), Central Andes (14), and Chocó (12) (Figure 6 – Source data 1). Some species also occur in other areas, such as the Atlantic Forest (14 spp.), Cerrado (12). Very few species are distributed across temperate latitudes: we only identified 8 species “Elsewhere”, 5 species occur in the Chaco region, and none in temperate South America. This distribution pattern is the most common in our mammal dataset (Figure 6 – Source data 1 and 2). Therefore, we think the contrasted diversification dynamics found here might be real, reflecting a better ability of New World monkeys to adapt to the changing conditions that occurred in South America during the Cenozoic, or be an artifact of taxonomic over-splitting of species in this clade, given that previous studies have suggested that the number of species has traditionally been inflated in this clade (see Springer et al. 2012 – PLoS One for a discussion on this taxonomic issue).

L384-386: It might be useful to note here that while this may be the case, terrestrial tetrapods haven't gone completely extinct, as they've persisted since the Carboniferous.

We agree, and we did not mean that terrestrial tetrapods have gone extinct but rather that they have experienced some periods of slowdown in diversification rates, or even periods of diversity decline (where the number of species has decreased) but eventually bounced back. We have clarified this sentence as follows: “Our results lend support to an alternative explanation for diversification slowdowns: the idea that tetrapods, for some periods, were less successful in keeping pace with a changing environment.”

L402-408: I like that attention is paid to potential biotic drivers of plant diversification; however, I'm also left wondering what role the evolution of angiosperm-dominated rainforests (which changed both the climate and vegetation structure) played any role in shaping the evolution of terrestrial tetrapods. I know it is beyond the scope of this paper to formally address it, but I think briefly mentioning the plausibility of this as a potential explanation for time-variable diversification dynamics would be beneficial.

This is a relevant point. We acknowledge that the evolution of angiosperm-dominated rainforests played an important macroevolutionary role in shaping the evolution of terrestrial biodiversity. Some of us have studied its role over the diversification of conifers, which have likely been outcompeted by angiosperms (e.g. Condamine et al. 2020 – PNAS). However, we don't think the evolution of an angiosperm-dominated forest significantly affected diversification dynamics of tetrapods in our study. It is considered that angiosperm-dominated forests were already established in the Neotropics by the Paleocene (Johnson and Ellis 2002 – Science; Wing et al. 2009 – PNAS; Carvalho et al. 2021 – Science), while the age of origin for most clades in our study postdates this period (Figure 2). It is possible that this biome transition affected the early evolution of a few clades of frogs and squamates, whose crown ages date back to the end of the Cretaceous. However, most cladogenetic events on these clades postdate the Paleocene. In the revised manuscript, we have incorporated the fact that our model design is limited to the tested hypotheses and that there could have been other untested variables, such as the role of angiosperm-dominated rainforests, that could drive Neotropical diversification: “Similarly, we did not assess the role of the emergence of angiosperm-dominated rainforests in the evolution of tetrapods. Angiosperm-dominated forests were already established in the Neotropics by the Palaeocene (104), while the age of origin for most clades in our study postdates this period (Figure 2)”.

L411-412: Any idea if this is due to the absence of other large clades restricted to the Neotropics, a lack of phylogenies for these, or because neotropical species are embedded in clades that have substantially radiated elsewhere (or some combination of these)?

An interesting point that is somewhat difficult to address. If we have to pick one of the three proposals, we would bet on the incomplete representation in phylogenies of species from large plant genera or from subclades within those genera. Most megadiverse plant genera (> 500 spp.) are well represented in the Neotropics (e.g., Anthurium, Astragalus, Begonia, Carex, Croton, Epidendrum, Eugenia, Lepanthes, Masdevallia, Maxillaria, Miconia, Mimosa, Myrcia, Passiflora, Peperomia, Piper, Pleurothallis, Psychotria, Salvia, Solanum, Tillandsia; Ulloa et al. 2017 – Science). Although there are phylogenetic studies including species from all of them, the sampling is largely incomplete in most cases. In general, phylogenetic studies of large groups can only afford to include a small fraction of the species, although they try to represent the morphological/geographical diversity of the group. A few examples of groups we are familiar with that still suffer from incomplete phylogenetic knowledge in the Neotropics are: Brongniartia, Dalbergia, Harpalyce, Zapoteca (Leguminosae); Acalypha, Croton sect. Adenophylli, Jatropha, Mabea, Sapium, Sebastiania (Euphorbiaceae); Eugenia, Myrcia, Myrcianthes (Myrtaceae); Arachnothryx, Palicourea, Psychotria (Rubiaceae); Ficus (Moraceae), Navia (Bromeliaceae); Abolboda, Xyris (Xyridaceae), Paepalanthus (Eriocaulaceae); Epidendrum (Orchidaceae).

Another issue to consider is that we are missing data from phylogenetic studies that were published after the closure of our data collection (2018–2022). It is inevitable to have a gap in the sampling due to the time lapse since data gathering and the actual publication of the study.

Finally, it is true that there are Neotropical species/groups that are embedded in larger clades, which are more species-rich elsewhere. For these cases, it is not possible nor relevant to study the diversification history in the Neotropics, because the Neotropical component in those groups is represented by a handful of species.

L438: I think it's important to remember that these biomes, ecoregions and biogeographic units may be relatively recent.

We agree. This is a very good point. Thank you for bringing it up. There is a body of evidence for the age of some of these biomes, which indeed indicates a recent age for some. For instance, the Cerrado is considered to have originated 9 million years ago (Simon et al. 2009 – PNAS). The Chocó region is supposed to be of more recent origin (Pérez-Escobar et al. 2019 – Front. Plant Sci.). This is relevant to explain why traditional bio-regionalization schemes are not appropriate to describe geographic patterns at the scale of this study. In the revised manuscript, we have mentioned this relevant point: “Often, these bioregions have been shown to be useful for categorizing actual species ranges, but they are less appropriate for examining endemism at the macroevolutionary scale. The age of origin of many of these bioregions postdates the origin of many of our clades (Figure 2). For instance, the Cerrado originated during the late Miocene (110), and the Chocó during the Pliocene-Pleistocene (111). In addition, most clades in our study appear distributed in most Neotropical ecoregions (Figure 6 – Source Data 1)”.

L449-452: How can species richness patterns identify shared evolutionary histories? It feels as though it might be challenging to disentangle whether these richness patterns reflect in situ diversification in that biome or dispersal into that biome.

The biogeographic clusters or macroregions identified here reflect areas of endemism at the clade level. These regions have been identified based on different clades sharing a similar pattern, in which most of the clade's species occur within the macroregion, and very few in other regions. This suggests that the contribution of in situ diversification might be more relevant than dispersal to explain the biotic assemblage of these macroregions. Accordingly, we initially defined these macroregions as regions of "shared evolutionary history". However, we realize that this term can be confusing and in the revised version of the manuscript we have rephrased it to clarify the meaning of these macroregions: “We propose here an alternative bioregionalization scheme of the Neotropical region that accounts for long-term regional assemblages at macroevolutionary scales (Figure 6). We identify five biogeographic units that represent macroregions where different independent Neotropical radiations occurred over millions of years of biotic evolution. These regions are defined in terms of species-richness patterns within clades (Figure 6 – Source Data 1; Figure 6 – Source Data 2), showing that species-rich clades in Amazonia, also tend to be species-rich in the Andes, Chocó, Guiana Shield, and Mesoamerica (biogeographic cluster 1), without excluding that some species within these clades occur in other regions…[]…As such, these biogeographic clusters form distinctive units of Neotropical evolution and represent long-term clade endemism. Within each of these clusters, the contribution of in situ diversification is therefore more relevant than dispersion to explain their biotic assemblage”.

L524-531: Please clarify if any occurrence (including a single occurrence when 99% of occurrences are from outside the neotropics) was enough to code the taxon as occurring in the Neotropics, or if a higher threshold was set (xx% of occurrences from Neotropics). And the same request for assigning to taxa to the 13 WWF biomes.

Yes, in our approach, we considered that a single occurrence was enough to assign a taxon to a given region. We are aware that this approach has limitations, especially since GBIF data often present inaccurate or erroneous occurrences (Maldonado et al. 2015 – Glob. Ecol. Biogeogr.). However, it is paramount to notice that, from the raw occurrence dataset, we removed occurrences with precision below 100 km, entries with mismatched georeference and country, duplicates, points representing country capitals or centroids.

By accepting all occurrences in our database, we would increase the probability to include erroneous records in our analyses, which can lead to an overestimation of species’ ranges. Conversely, applying a given threshold to code a taxon as occurring in an area could lead to the opposite result: under-estimation of species’ ranges if the distribution of the species is poorly known or if some regions are less well sampled than others, which is often the case for Neotropical taxa. Some Neotropical regions accumulate the greatest number of missing or undescribed species, and the lowest biodiversity monitoring efforts (Kier et al. 2005 – J. Biogeogr.; Pimm and Joppa 2015 – Ann. Missouri Bot. Gard.). For example, the southern section of the Amazon basin and northern Colombia, are presumably the most species-rich of all data gaps in the world for plants (Frodin 2001 Cambridge University Press; Perez-Escobar et al. 2022 – TIPS). We considered that no methodological option came without problems.

Given the limited knowledge on the distribution of many Neotropical species, we chose to follow the first approach. However, in order to minimize the impact of this choice, we manually inspected the distribution of a number of records, especially when species were assigned to more than >2 areas. Specifically, we surveyed the distribution of 4,477 species of the 12,512 species included in our dataset, checking for completeness and accuracy with reference to different databases (AmphibiaWeb 2018, Uetz et al. 2018, GBIF.org 2018, IUCN 2018, etc.). Most species in our dataset occurred in a single WWF ecoregion (7,334; please see the table below), reflecting a well-known pattern in the Neotropics of a large proportion of assemblages of endemic species with small ranges (Pimm et al. 2014 – Science). Only 5,178 species occur in more than 1 area. We increased the surveying effort with the number of areas; for comparison, we checked the distribution of 26% of the single area species, and almost 50% of the species with distributions in more than 1 area (from 35% of the species occurring in 2 areas to 100% of the species occurring in 13 areas). We are painfully aware that even with this effort it is still possible that there may be errors in our database, but we still hope that general patterns could be extracted from this data.

Author response table 1 shows the number (Nb.) of species assigned to 1 to 13 WWF ecoregions, and the proportion of species for which we checked their distribution.

Author response table 1
Nb. WWF areasNb. speciesNb. species checked% species checked
17,3341,91826.15
22,60992235.34
31,33952939.51
468737053.86
530721570.03
621818484.40
712710381.10
81119181.98
9776787.01
10463576.09
11323093.75
121111100
1366100

L685-687: Was the 60% criterion applied above used here to determine whether clades were scored as mixed, lowland, montane, or highland?

Unfortunately, we were unable to follow the same quantitative classification criteria for coding the main elevation of each clade in our dataset as done for their distribution. Many of the 12,512 species in our 150 phylogenies lacked accurate altitudinal data, as GBIF georeferenced records often come without this information. For example, we estimated that only 40% of the mammal occurrences downloaded in our dataset included elevation data; 34% of the plants. Surprisingly, occurrence data for birds only included associated elevation information in 4% of cases. We therefore considered the altitudinal range of each species could be incompletely described based on the sparse occurrence data available for many species in our dataset, and chose to rely on the descriptions in the literature, keeping in mind that groups described with most of their species in a given elevation would be assigned to this elevation (mimicking the 60% criterion used for the distribution above). We have clarified our approach in the revised version as follows: “We additionally classified clades according to the main elevational range of their constituent species following literature descriptions rather than a purely quantitative approach as for the distribution above, because GBIF records in our dataset often came without associated altitude data (<30%):”

L699-700: Doesn't this show that there is NOT phylogenetic signal (as it is written, it states that any trait contains phylogenetic signal, even though the K values are low and non-significant).

Thanks for noticing this. As suggested, it is in fact the opposite. We have corrected this sentence by replacing “any” by “no”.

[Editors’ note: The authors appealed the decision. What follows is the authors’ response to the second round of review.]

Both Reviewers and we (editors) valued the work done in the revision, but further comments raised by Reviewer 1 further weakened the validity of your approach. Specifically, the lack of a controlled methodology (by confirmatory or accuracy-testing simulations) made it difficult to evaluate if any strong bias exists in your results, or if your datasets have enough power to sustain your claims.

Reviewer #1 (Recommendations for the authors):

I previously reviewed this manuscript in December of 2021. I had several concerns. One was that the main method used to estimate diversification over time might be problematic.

We thank you for your insightful comments and review. We apologize for not being clear enough in our previous responses to those comments. We have now provided additional information and data, and have included a new section in the manuscript discussing in detail the limitations of the study. We hope that the responses given below and corresponding changes in the manuscript will help clarify these issues.

As previously articulated, we feel this study should not be penalized by the current lack of scientific consensus on a topic that remain central to modern studies in evolutionary biology: the estimation of diversification rates based on molecular phylogenies and time-variable homogeneous birth-death ‘BD’ models. Debate on this subject is intense (Louca and Pennell 2020; Helmstetter et al. 2022; Morlon et al. 2022; Höhna et al. 2022) and is unlikely to be solved in the short term. Our paper does not neglect the debate. By comparing our results with pulled diversification models (supposed to correct for the identifiability issue raised by recent studies associated with traditional diversification rates) as far as these two estimates could be compared, and discussing the limitations of our results, we would argue that we have done our very best efforts to present the limits and extension of our conclusions. We also think that eLife is a key journal for publishing results that bring discussion further, like those in this manuscript, rather than avoid topics for which consensus has not yet been reached.

Another was that there was little comparison among different regions to understand patterns of diversity among regions in the Neotropics. I appreciate that the authors have made efforts to address these concerns, but I think that these problems remain somewhat problematic.

We have addressed this comment below, and hopefully you will find this time our answers more satisfactory. However, we think that our points of view and preferred methodology may differ on how biogeographic regions should be delimited and diversification rates analyzed. We think such differential preferences are part of a healthy scientific arena and should be allowed expression, since none of them present logical or obvious flaws.

First, despite what the authors claim, I think the main approach used has not really been thoroughly tested with simulations. The authors claim that it has been, specifically by Morlon et al. (2011), Lewitus and Morlon (2018), and Condamine et al. (2019). The paper by Morlon et al. (2011) contains simulations, but (as far as I know) those simulations do not actually address the ability of the method to accurately choose among different models of diversification over time.

We respectfully disagree on this point. We acknowledge that not all possible simulation scenarios have been performed (there are simply too many cases). But compared to other birth-death models, the RPANDA models have been thoroughly tested with simulations. Time-dependent birthdeath models are not only able to recover (1) the parameter values (including negative net diversification rates, Morlon et al. 2011), but these models are also able to (2) correctly detect shifts of diversification (with regularization techniques as proposed in Morlon et al. 2022), (3) recover the good diversification model, and (4) infer the correct paleodiversity dynamic by taking into account clade, clade diversity and rates (Mazet et al. 2022: https://www.biorxiv.org/content/10.1101/2022.05.10.490920v1).

The paper by Condamine et al. (2019) does not contain simulations at all.

We agree that this study does not include true simulations and have therefore excluded this citation from references about simulations, and refer instead to the Lewitus and Morlon (2018)'s paper which did so (see below our response to your comment related to this paper). In the Condamine et al. (2019)'s paper, the authors have shown that the specificities of the environmental curve matter for recovering a statistical support for a temperature-dependent model.

The paper by Lewitus and Morlon (2018) contains simulations, but primarily compares the ability of the method to distinguish time-dependent and temperature-dependent models (their Figure 3). But it is unclear if these two classes of models can be accurately distinguished from cases where the true model is a constant-rate model. Furthermore, looking at Figure 3 of Lewitus and Morlon (2018) it can be seen that this method often selects a constant-rate model when the true model is temperature dependent, if the effect of temperature dependence on diversification over time is not strong enough. The authors include in this list only papers co-authored by the developer of the method, and not other papers that found the method to be more problematic.

We acknowledge that this method has limitations (as all methods do). Over the last decade, the RPANDA package has incorporated several birth-death models that are now made to be compared in a common statistical framework, which is not the case if we perform other birth-death models like BAMM, RevBayes or SSE-based models. These latter models will be useful to address other questions like the role of traits on diversification or the presence of rate heterogeneity across the studied phylogenies. As we focus on drivers of diversification, there are not many models we can use to test the hypotheses we proposed in the Introduction. Also, we aim at a clear story and do not want to perform a battery of analyses that would make the study long and more confusing for the readers.

We think it's an important point that time-dependent and temperature-dependent models are distinguishable; it’s not the case between time- and diversity-dependent models (Pannetier et al. 2021). Importantly, temperature-dependent models are also able to accurately recover simulated parameter values (even for the extinction rate). According to simulations, the model selection tends to be sensitive when the dependency parameter [α] ranges between -0.1 and 0.1. This means that we are conservative when we conclude that temperature-dependent models are estimated as best fitting.

We would like to point out that we have calculated the proportion of phylogenies whose diversification rates are best explained by a model with constant, time-dependent, temperaturedependent, or uplift-dependent diversification based on (1) the best fitting model, and (2) the secondbest model (Figure-5—figure supplement 1). If the constant-rate model overfits temperaturedependent trees, then we would expect that the temperature-dependent models are the second bestfit model in most cases. We find it's not the case in our analyses. Of the 76 clades that have a constantrate model as the best fit, 50% (38/76) have temperature-dependent models as the second best fit, 40% (30/76) have time-dependent models, and 10% (8/76) have Andean-dependent models. Please see our new Figure-5-source-data-2. This suggests that there is no major bias in our model selection. These results can even suggest there would be a similar bias towards time-dependent models. Furthermore, when evaluating the α values of those clades which are best fit by a constant-rate model and second best fit by a temperature-dependent model, we find that only 26% (10/38) of the temperature models have α values ranging between -0.1 and 0.1, which according to simulations is the value range where model selection tends to be sensitive. (Please see Author response image 3 and our new Figure-5-source-data-2). This number gets reduced to 6% when considering the whole sample size (10/150), meaning that only 6% of the trees in our dataset are susceptible to suffer from this potential bias (i.e. select a constantrate models when the true model is a temperature-dependent model).

Author response image 3
Α value of the clades which are best fit by a constant-rate model and second best fit by a temperature-dependent model.

In red those values ranging between -0.1 and 0.1. Eleven trees have constant speciation (α does not vary) and thus they are not represented here.

We have added this information (model sensitivity and evaluation of our data) to a new section ("Limitations and perspectives") in the Discussion. We now discuss these limitations of the RPANDA approach on this section.

It is also notable that the diversification dynamics are not significantly different between plants and animals when using PDR (Table 2). Therefore, not all of the main results are concordant between these two methods.

We may not have been sufficiently clear in reporting these results. In fact, Table 2 does show that pulled diversification rates (PDR) are significantly different between plants and animals (P < 0.00), in agreement with results based on traditional BD models. Diversification trajectories, derived from these rates, are not significantly different in terms of proportion of clades experiencing expanding vs. declining speciation trends across taxonomic groups. Unfortunately, inferences on speciation trends are not sufficient to derive net diversification dynamics and thus to support/reject results based on BD rates. Extinction cannot be inferred based on PDR, only speciation (Louca and Pennel 2020), but estimates of extinction are needed to derive net diversification dynamics (net diversification = speciation minus extinction), as we did base on BD models. For example, we cannot rule out that the decreases in speciation in plants based on PDR (Figure 3) were accompanied by larger decreases in extinction. Under this scenario, net diversification rates could still increase through time, and we could still recover similar expanding diversification dynamics for plants as we found based on BD models. Thus, direct comparisons between traditional BD and PDR models were not possible in this case.

We explain in the Discussion that results from PDR and BD are not directly comparable on this point, but we realize this might have been confusing or unclear in the previous version. We have therefore clarified it further in the text. It now reads: “The study of PDR did not help to confirm/reject these conclusions. Rates from PDR models are significantly different between plants and animals (p<0.00), in agreement with results based on traditional models (Table 2). Diversification trajectories derived from these rates are not different, with plants and animals showing an equivalent fraction of phylogenies showing a decay of speciation (Figure 3b). Since extinction dynamics cannot be derived from PDR models, we do not know if speciation slowdowns detected in plants based on PDR were accompanied by larger extinction declines (Figure 3). Thus, we cannot rule out the scenario of expanding dynamics (Sc. 1, 2) for plants found based on traditional birth-death models.”

I found the comparison among biogeographic regions to be unsatisfying overall, because the regions are extremely coarse grained and each incorporates so many different regions.

We understand why the reviewer finds the biogeographic regions identified here as being broad in comparison with traditional subdivisions of the Neotropics. We also agree with the principle of using the maximum possible resolution afforded by data availability, and agree that "finer" regions would be desirable if our evolutionary units of analysis were species – as has been done in some previous studies comparable to ours. In contrast, our study is novel in using clades (lineages that radiated in the Neotropics) as evolutionary units for the spatial analyses.

While broad biogeographic regions would not be meaningful when working at the species level, they are adequate when working at the clade-level scale. In fact, we initially tried a finer definition of Neotropical regions, but it did not work for our data. Figure-6-source-data-1 shows that finer regions, such as ecoregions (Olson, 2001), do not represent patterns of endemism for our clades, i.e., most of the clades have species distributed in most ecoregions, and almost none is endemic to one ecoregion. Another important point against fine-scale biogeographic regions is that the origin of many of the traditional Neotropical subdivisions postdates the origin of many of the studied clades.

For example, the Cerrado formed in the late Miocene (Simon et al., 2009), and the Chocó in the Pliocene-Pleistocene (Pérez-Escobar et al., 2019), which stresses the conclusion that finer bioregions were not yet formed when most clades originated (Figure 2).

We would like to stress that the definition of our biogeographic regions is the result of a quantitative analysis, rather than resulting from any a priori delimitation based on expert’s knowledge. Our regions derive from the analysis of patterns of endemicity in our sample of 150 Neotropical clades. Importantly, these regions are overall congruent with the biogeographic sub-regionalization of the Neotropics proposed in previous influential studies (Morrone 2017, 2022).

We believe that the use of clades (rather than species) as evolutionary units in our biogeographic comparisons is a novelty of our study and has several advantages. We were able to compare diversification trends through time (i.e. constant, expanding, declining) across regions, and not just present-day diversification rates as in previous studies (Harvey et al., 2020 – Science; Quintero and Jetz 2018 – Nature; Smith et al., 2014 – Nature). The later approach is limited because when rates vary through time, diversification could be higher in one region than in another without providing information on the underlying diversification dynamic through time. We think the existence of regional arenas of diversification in the Neotropics at the clade level and at deep time scales is a key finding of this study. We have extended the discussion to explain this: “The use of clades (rather than species) as evolutionary units in our biogeographic comparisons is original, and allowed us to compare diversification trends through time (i.e. constant, expanding, declining) across regions, and not just present-day diversification rates, as in different comparable studies (e.g., Harvey et al., 2020; Quintero and Jetz 2018; Smith et al., 2014). Present-day diversification rates are structured geographically in the Neotropics (Harvey 2020, Jetz 2012, Quintero and Jetz 2018; Rangel 2018), but our study shows that present diversification do not represent long-term evolutionary dynamics. The lack of a clear geographic structure of long-term diversification suggests that the evolutionary forces driving diversity in the Neotropics acted at a continental scale when evaluated over tens of millions of years. Evolutionary time and extinction could have eventually acted as levelling agents of diversification across the Neotropics over time.”

Furthermore, the comparison of 97 clades in one region (Mesoamerica+Amazonia+Andes and others) vs. only 10 clades in another (southern South America) also seems unsatisfying.

We agree with the reviewer. Our conclusions derive from the study of a fraction of the Neotropical diversity, where tropical rainforest lineages from the broad “pan-Amazonian” region are most abundant. Although sample size in our study is large (150 observations), some categories of these variables are poorly represented, which might limit the performance of some statistical tests. For instance, there are 97 phylogenies distributed in the biogeographic cluster 1, while only 10 in cluster 2, 4 in cluster 3. Note that there are other clades (39) containing species on poorly represented regions that fall in the “mixed” category, as they share species with different areas. We argue, however, that our sampling does include representatives from all the main regions in the Neotropics. Yet, we did not identify a common diversification trajectory among the fewer clades distributed on poorly represented regions (e.g. southern South America clades experienced all gradual, exponential, and declining dynamics, as did the clades from other regions; Figure 6). Furthermore, it is reasonable to assume that our sampling reflects a fair proportion of species’ distributions in the Neotropics considering the extension of these regions in the Neotropics, and the representativeness of our dataset; at least for tetrapods, it includes ~60% of all described species. We did not choose our clades a priori for belonging to one area or the other, and our sampling probably represents the distribution of species on these regions. Furthermore, the hypothesis of comparable diversification gains support when comparing raw diversification rate estimates (Figure 6d,e), and not just their derived species richness trends (Sc. 1–4; Figure 6f,g). Still, future studies with larger sample sizes per region will be needed to clarify the generality of our results. We have expanded on this issue in the new section ("Limitations and perspectives") in the Discussion.

I appreciate that the authors describe their four scenarios explicitly in the Introduction, but I do not think that knowing the frequency of these four scenarios is "required to understand the origin and maintenance of Neotropical diversity" (line 112).

It is true that describing the frequency of these four scenarios on the dynamics of Neotropical diversity over the Cenozoic will not resolve "the origin and maintenance of Neotropical diversity", but we think that this information can help to shed light on it. We have rephrased this statement in the Introduction, which now reads: “Yet, such an assessment would contribute to understand the origin and maintenance of Neotropical diversity.”

Also, I do not see why the accumulation of species through "pulses" corresponds to the exponential expansion scenario. Should not constant speciation and low, constant extinction lead to an exponential increase in diversity over time? Even in Figure 1, I do not see the pulses in Scenario 2. A "pulse" implies a pattern that is discontinuous or episodic over time.

The reviewer is right. The term “pulses of speciation” refers to episodic discontinuous events and does not accurately describe the exponential increase of diversity described in Scenario 2. We have corrected the text, which now reads: “An exponential increase in diversity model asserts that species richness accumulated faster towards the present”. Note that a scenario of constant speciation and low constant extinction will not lead to an exponential increase in diversity over time, but to a linear increase (see scenario on Figure 1b).

The finding that diversification patterns are different between plants and animals is interesting, but not entirely novel (see for example, Hernandez-Hernandez et al. 2021; Biological Reviews on the faster rates of diversification in plants vs. animals). It is also notable (again) that the diversification dynamics over time are not significantly different between plants and animals when using PDR, even if the rates are.

Thank you for this relevant reference. We have included this reference in the discussion. Now reads: “Net diversification rates were also higher in plants (Figure 8), in agreement with previous studies (Hernandez-Hernandez et al., 2021)”. Regarding the apparent contradiction between the PDR and BD results, please see our previous reply.

Finally, I am not sure how relevant these results are to future climate change, as is stated in the Abstract. It seems like this idea is not addressed in the paper itself.

Thank you for this comment. We have rephrased the sentence as follows: “These opposite evolutionary patterns may reflect different capacities for plants and tetrapods to cope with past climate changes”.

Specific comments:

Lines 285 to 287: This needs to be rewritten since the results seem to contradict the conclusions.

Thank you for noticing this mistake. Following the suggestion of Reviewer 3 we have corrected the sentence as follows: “we find that no continuous or multi-categorical trait shows a phylogenetic signal”.

Lines 340-342. The study cited focused on BiSSE-type models, not the approach used here.

The reviewer is right. Simulations in Davis et al. (2013) were conducted using BiSSE-type models. However, sample size limitations are not particular of birth-death models, but to any kind of model. We have changed this sentence as follows: “Alternatively, the power of birth–death models to detect rate variation decreases with the number of species in a phylogeny, as shown with different diversification approaches (Davis et al. 2013: Lewitus and Morlon 2018; Burin et al. 2019)”.

Line 374. Change "plant phylogenies are significantly worst sampled than those of most other tetrapods" to "plant phylogenies are significantly less sampled than are tetrapod phylogenies".

Thanks! Corrected.

Lines 415 to 416. How can there be microevolutionary studies of diversification?

The reviewer is right. We have reworded this paragraph and delete this sentence.

Lines 538 to 540. The paper by Kreft and Jetz does not address diversification, only present-day species richness.

Thanks for noticing this misquotation. We have removed this reference and replaced it by: Jetz et al., 2012; Harvey et al., 2020; Quintero and Jetz 2018; Rangel et al., 2018.

Lines 554 to 555. The "diverse range of diversification methods compared here" is actually only two diversification methods, by my count.

We have corrected this sentence. It now reads: “and the diversification models compared here”.

Line 563. What "substantial proportion of Neotropical diversity" is accounted for by these models?

The proportion of models accounted by saturated and declining models is 30%. If the reviewer agrees, we prefer not to include this information in the conclusions, as we do not think it is the appropriate place to put numbers. Besides, it is already presented in the abstract, results and discussion.

[Editors’ note: what follows is the authors’ response to the third round of review.]

In this revised version, we stress that you must carefully address the following: (1) Please be more cautious about reporting the constraints of your chosen methods, (2) Please dilute the claim that you are resolving South American diversification, (3) Please address the several aspects of the most recent reviews which were not addressed in your appeal letter, including (but not limited to): your reliance on one specific method to calculate diversification rates, an apparent lack of simulations to support your chosen method, the coarse-grained resolution of the study and imbalance in clade numbers investigated in different regions, the request for a clearer and more critical projection of scenarios resulting from the proposed speciation patterns (do pulses necessarily lead to exponential expansion), and being careful to base conclusions about diversification patterns on dynamics and rates.

Thank you very much for considering our appeal letter and for allowing us to submit a revised manuscript and a response to the reviewer's comments. We are also grateful for the suggestions of the three reviewers whose comments contributed to improve the paper. We have carefully addressed all the comments raised by the reviewers — please see our detailed answers below in blue font.

In summary, we agree with the suggestions put forward and have therefore toned down throughout the main text our claims that the results of our analyses are resolving the origin of the extraordinary South American diversity. Of course, such claim would be inappropriate, as our results can only contribute to our overall understanding of this complex and multifaceted topic. We have also tried to clarify what has and has not been done regarding simulations and birth-death modeling (please see our answer to reviewer 1 and our changes in the main text). We also discuss how, compared to other birth-death models, the RPANDA models have been thoroughly tested using simulations.

We acknowledge that all possible simulation scenarios have not been performed – there are simply too many possibilities. Previous studies, cited in the revised text, show that these models are not only able to recover the parameter values – including negative net diversification rates, which are often underestimated by models such as BAMM – but are also able to detect shifts of diversification, as well as the best-fitting diversification model. According to previous simulations (e.g., Lewitus and Morlon 2018), model selection tends to be sensitive to some particular scenarios. We now explicitly discuss such scenarios, and quantify the frequency with which they might potentially occur in our data (ca. 6% of our trees; please see our new Figure-5-source-data-2). This evaluation indicates that potential methodological biases are not significantly affecting our study, nor would have an effect on our conclusions.

We have now worked further to improve and clarify the reliability and limitations of our study by adding a specific section ("Limitations and perspectives") in the Discussion, in which we not only explain why birth-death models in RPANDA and diversification parameters are demonstrably identifiable, but also highlight that our results are based on traditional diversification rates and on the Pulled Diversification Rates method (which corrects for the identifiability issue raised by studies associated with traditional diversification rates).

Regarding the “coarse” definition of biogeographic areas raised by the reviewer, we have very carefully considered this feedback. We agree with the principle of using the maximum possible resolution afforded by data availability, and agree that "finer" regions would be desirable if our evolutionary units of analysis were species – as has been done in some previous studies comparable to ours. In contrast, our study is novel in using clades (lineages that radiated in the Neotropics) as evolutionary units for the spatial analyses. This means that finer areas cannot be used to categorize clade distributions for the simple reason that some of the finer regions used in previous studies conducted at the species level did not exist at the origin (and during the evolutionary history) of most of the clades.

As mentioned in the previous round of reviews, we actually attempted to apply finer regions and quickly realized that they did not produce biologically and biogeographically meaningful results. As shown in our Figure-6-source-data-1, finer regions, such as ecoregions (Olson, 2001), do not represent patterns of endemism for our clades, i.e., most of the clades have species distributed in most ecoregions, and almost none is endemic to one ecoregion. To analyze clade endemism and perform biogeographic comparisons of diversification at the clade level, we therefore needed to identify broader regions. We have done this following a quantitative analysis, which the reviewer has not found to be problematic, and not based on any a priori subjective decision. Moreover, our regions are overall congruent with the biogeographic sub-regionalization of the Neotropics proposed in influential works by Morrone (2017, 2022), which will make the results of our study more easily comparable with previous studies.

Although using clades (rather than species) as evolutionary units requires a larger spatial resolution for the biogeographic analyses, we think that our approach represents a major novelty of the study, bringing several advantages. Most importantly, we were able to compare diversification trends through time (i.e. constant, expanding, declining) across regions, and not just present-day diversification rates, as in comparable previous studies (e.g., Harvey et al., 2020 – Science; Quintero and Jetz 2018 – Nature; Smith et al., 2014 – Nature). A species-based approach is limited in scope because when rates vary through time, present diversification could be higher in one region than in another, without providing information on the underlying temporal dynamic. We think the existence of regional arenas of evolution in the Neotropics at the clade level and at large time scales is a key finding of this study and it should stimulate debate and further research. We have tried to clarify this in the reply to the reviewer and in the revised manuscript.

https://doi.org/10.7554/eLife.74503.sa2

Article and author information

Author details

  1. Andrea S Meseguer

    1. Institut des Sciences de l’Evolution de Montpellier (Université de Montpellier | CNRS | IRD | EPHE), Place Eugène Bataillon, Montpellier, France
    2. Real Jardín Botánico (RJB), CSIC, Madrid, Spain
    Contribution
    Conceptualization, Resources, Data curation, Software, Formal analysis, Investigation, Methodology, Writing – original draft, Writing – review and editing
    For correspondence
    asanchezmeseguer@gmail.com
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-0743-404X
  2. Alice Michel

    1. Institut des Sciences de l’Evolution de Montpellier (Université de Montpellier | CNRS | IRD | EPHE), Place Eugène Bataillon, Montpellier, France
    2. Department of Anthropology, University of California, Davis, United States
    Contribution
    Resources, Data curation, Software, Formal analysis, Writing – review and editing
    Competing interests
    No competing interests declared
  3. Pierre-Henri Fabre

    1. Institut des Sciences de l’Evolution de Montpellier (Université de Montpellier | CNRS | IRD | EPHE), Place Eugène Bataillon, Montpellier, France
    2. Mammal Section, The Natural History Museum, Cromwell Road, London, United Kingdom
    3. Institut Universitaire de France (IUF), Paris, France
    Contribution
    Resources, Data curation, Formal analysis, Investigation, Methodology, Writing – review and editing
    Competing interests
    No competing interests declared
  4. Oscar A Pérez Escobar

    Royal Botanic Gardens, Kew, United Kingdom
    Contribution
    Resources, Writing – review and editing
    Competing interests
    No competing interests declared
  5. Guillaume Chomicki

    Ecology and Evolutionary Biology, University of Sheffield, Sheffield, United Kingdom
    Contribution
    Resources, Writing – review and editing
    Competing interests
    No competing interests declared
  6. Ricarda Riina

    Real Jardín Botánico (RJB), CSIC, Madrid, Spain
    Contribution
    Resources, Data curation, Investigation, Writing – review and editing
    Competing interests
    No competing interests declared
  7. Alexandre Antonelli

    1. Royal Botanic Gardens, Kew, United Kingdom
    2. Gothenburg Global Biodiversity Centre, Department of Biological and Environmental Sciences, University of Gothenburg, Gothenburg, Sweden
    3. Department of Plant Sciences, University of Oxford, Oxford, United Kingdom
    Contribution
    Resources, Writing – review and editing
    Competing interests
    No competing interests declared
  8. Pierre-Olivier Antoine

    Institut des Sciences de l’Evolution de Montpellier (Université de Montpellier | CNRS | IRD | EPHE), Place Eugène Bataillon, Montpellier, France
    Contribution
    Resources, Funding acquisition, Writing – review and editing
    Competing interests
    No competing interests declared
  9. Frédéric Delsuc

    Institut des Sciences de l’Evolution de Montpellier (Université de Montpellier | CNRS | IRD | EPHE), Place Eugène Bataillon, Montpellier, France
    Contribution
    Conceptualization, Funding acquisition, Project administration, Writing – review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-6501-6287
  10. Fabien L Condamine

    Institut des Sciences de l’Evolution de Montpellier (Université de Montpellier | CNRS | IRD | EPHE), Place Eugène Bataillon, Montpellier, France
    Contribution
    Conceptualization, Resources, Data curation, Software, Formal analysis, Funding acquisition, Writing – original draft, Project administration, Writing – review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-1673-9910

Funding

Agence Nationale de la Recherche (ANR-10-LABX-25-01)

  • Andrea S Meseguer
  • Alice Michel
  • Pierre-Henri Fabre
  • Pierre-Olivier Antoine
  • Frédéric Delsuc
  • Fabien L Condamine

Agence Nationale de la Recherche (ANR-17-CE31-0009)

  • Pierre-Henri Fabre
  • Pierre-Olivier Antoine
  • Frédéric Delsuc
  • Fabien L Condamine

Ministerio de Ciencia e Innovación (PID2020-120145GA-I00)

  • Andrea S Meseguer

Comunidad Autonoma de Madrid, Atraccion de Talento (2019-T1/AMB-12648)

  • Andrea S Meseguer

Ministerio de Ciencia e Innovación (PID2019-108109GB-I00)

  • Ricarda Riina

Swedish Research Council (2019-05191)

  • Alexandre Antonelli

Natural Environment Research Council (NE/S014470/1)

  • Guillaume Chomicki

Swiss Orchid Foundation

  • Oscar A Pérez Escobar

Ministerio de Ciencia e Innovación (IJCI-2017-32301)

  • Andrea S Meseguer

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

We are grateful to the three reviewers and the editors, for many insightful comments that helped improve the quality of our work. We thank all researchers who shared their published data through databases or with us directly (Drs Arevalo, Martins, Fortes Santos, Simon, Lohmann, Mendoza, Swenson, Erkens, van der Meijden, and Freitas). Drs J Muñoz, J Lobo, I Sanmartín, S Louca, P Manzano, and M Godefroid for invaluable comments on the manuscript and/or analyses. This work was funded by an 'Investissements d’Avenir' grant managed by the Agence Nationale de la Recherche (CEBA, ref ANR-10-LABX-25-01) and the ANR GAARAnti project (ANR-17-CE31-0009). ASM was also supported by the Atracción de Talento CAM program (2019-T1/AMB-12648), Juan de la Cierva grant (IJCI-2017-32301), and Grant PID2020-120145GA-I00 funded by MCIN/AEI/ 10.13039/501100011033 and by the European Union. AA is supported by the Swedish Research Council (2019-05191), and the Royal Botanic Gardens, Kew. OAPE is funded by the Swiss Orchid Foundation and the Sainsbury Orchid Fellowship at the Royal Botanic Gardens, Kew. GC is funded by a Natural Environment Research Council Independent Research Fellowship (NE/S014470/1). RR is funded by the Spanish Ministry of Science (PID2019-108109GB-I00, AEI/FEDER). This is contribution ISEM 2022-270 of the Institut des Sciences de l’Evolution de Montpellier.

Senior Editor

  1. Meredith C Schuman, University of Zurich, Switzerland

Reviewing Editor

  1. David A Donoso, Escuela Politécnica Nacional, Ecuador

Reviewer

  1. Yaowu Xing, Xishuangbanna Tropical Botanical Garden, China

Publication history

  1. Preprint posted: February 25, 2021 (view preprint)
  2. Received: October 7, 2021
  3. Accepted: October 26, 2022
  4. Accepted Manuscript published: October 27, 2022 (version 1)
  5. Accepted Manuscript updated: October 28, 2022 (version 2)
  6. Version of Record published: November 16, 2022 (version 3)

Copyright

© 2022, Meseguer et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 867
    Page views
  • 302
    Downloads
  • 0
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Open citations (links to open the citations from this article in various online reference manager services)

Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)

  1. Andrea S Meseguer
  2. Alice Michel
  3. Pierre-Henri Fabre
  4. Oscar A Pérez Escobar
  5. Guillaume Chomicki
  6. Ricarda Riina
  7. Alexandre Antonelli
  8. Pierre-Olivier Antoine
  9. Frédéric Delsuc
  10. Fabien L Condamine
(2022)
Diversification dynamics in the Neotropics through time, clades, and biogeographic regions
eLife 11:e74503.
https://doi.org/10.7554/eLife.74503
  1. Further reading

Further reading

    1. Evolutionary Biology
    2. Genetics and Genomics
    Yi Feng, Rafik Neme ... Laura F Landweber
    Research Article

    Ciliates are microbial eukaryotes that undergo extensive programmed genome rearrangement, a natural genome editing process that converts long germline chromosomes into smaller gene-rich somatic chromosomes. Three well-studied ciliates include Oxytricha trifallax, Tetrahymena thermophila and Paramecium tetraurelia, but only the Oxytricha lineage has a massively scrambled genome, whose assembly during development requires hundreds of thousands of precise programmed DNA joining events, representing the most complex genome dynamics of any known organism. Here we study the emergence of such complex genomes by examining the origin and evolution of discontinuous and scrambled genes in the Oxytricha lineage. This study compares six genomes from three species, the germline and somatic genomes for Euplotes woodruffi, Tetmemena sp., and the model ciliate Oxytricha trifallax. To complement existing data, we sequenced, assembled and annotated the germline and somatic genomes of Euplotes woodruffi, which provides an outgroup, and the germline genome of Tetmemena sp.. We find that the germline genome of Tetmemena is as massively scrambled and interrupted as Oxytricha's : 13.6% of its gene loci require programmed translocations and/or inversions, with some genes requiring hundreds of precise gene editing events during development. This study revealed that the earlier-diverged spirotrich, E. woodruffi, also has a scrambled genome, but only roughly half as many loci (7.3%) are scrambled. Furthermore, its scrambled genes are less complex, together supporting the position of Euplotes as a possible evolutionary intermediate in this lineage, in the process of accumulating complex evolutionary genome rearrangements, all of which require extensive repair to assemble functional coding regions. Comparative analysis also reveals that scrambled loci are often associated with local duplications, supporting a gradual model for the origin of complex, scrambled genomes via many small events of DNA duplication and decay.

    1. Evolutionary Biology
    2. Microbiology and Infectious Disease
    Vishhvaan Gopalakrishnan, Dena Crozier ... Jacob G Scott
    Feature Article Updated

    A morbidostat is a bioreactor that uses antibiotics to control the growth of bacteria, making it well-suited for studying the evolution of antibiotic resistance. However, morbidostats are often too expensive to be used in educational settings. Here we present a low-cost morbidostat called the EVolutionary biorEactor (EVE) that can be built by students with minimal engineering and programming experience. We describe how we validated EVE in a real classroom setting by evolving replicate Escherichia coli populations under chloramphenicol challenge, thereby enabling students to learn about bacterial growth and antibiotic resistance.