What drives diversification in a pantropical plant lineage with extraordinary capacity for long‐distance dispersal and colonization?

Colonization of new areas may entail shifts in diversification rates linked to biogeographical movement (dispersification), which may involve niche evolution if species were not exapted to new environments. Scleria (Cyperaceae) includes c. 250 species and has a pantropical distribution suggesting an extraordinary capacity for long‐distance dispersal and colonization. We investigate patterns of diversification in Scleria, and whether they are coupled with colonization events, climate niche shifts or both.


| INTRODUC TI ON
Clade-specific bursts in diversification rates have been associated with morphological changes termed key innovations related to novel niche invasions, increasing individual fitness or reproductive isolation (e.g. Naciri & Linder, 2020). Dispersification or 'key opportunity' has been defined as shifts in diversification rates associated with biogeographical movements without necessarily invoking key innovations (Donoghue & Sanderson, 2015;Moore & Donoghue, 2007;. Dispersification and key innovations do not exclude each other and shifts in diversification rates may be related to synergetic action of both (Moore & Donoghue, 2007).
After colonization, (a)biotic conditions may limit establishment of species. If species are unable to adapt to new environmental conditions (i.e. niche conservatism), they may only colonize environments similar to those in their original range (Pulliam, 2000). However, species can sometimes adapt to new environmental conditions (i.e. niche shifts, niche evolution), enabling colonization of new (a)biotic conditions (e.g. Pearman, Guisan, Broennimann, & Randin, 2008;Villaverde, González-Moreno, Rodríguez-Sánchez, & Escudero, 2017). A key innovation may allow a species to invade a new niche (novel niche invasion) allowing a subsequent shift in diversification rates (Boucher et al., 2012). In a scenario of dispersification (e.g. Donoghue, 2008), new environments reached after dispersal will often be colonized with lineages from environmentally similar areas (Moore & Donoghue, 2007). Alternatively, shifts in diversification rates have been associated with major historical climatic events (Erwin, 2009).
A family-wide biogeographical study of the sedge family Cyperaceae has shown that species-rich sedge clades are more widespread, occupy more niche space and diversify faster than species-poor lineages (Spalink et al., 2016). Here, we focus on a species-rich pantropical sedge clade, the genus Scleria P.J. Bergius Most Scleria species occur in the tropical zone below 1,600 m, with some extending into warm temperate regions (Bauters, 2018). In the tropics, species growing at higher elevations are also adapted to a more temperate climate than species growing in lowlands.
Approximately 112 species are known from the Americas, 105 from Africa and 58 from Asia and Oceania. The pantropical distribution of Scleria suggests an extraordinary capacity for long-distance dispersal (LDD) and colonization. The dispersal unit of Scleria, that is the nutlet (plus hypogynium in most groups), are dispersed by different vectors. Abiotic dispersal via gravity and/or wind is common in species with unspecialized nutlets. Species of section Ophryoscleria have a corky-swollen cupule which stays attached to the nutlet, making it buoyant (Robinson, 1962). These species occur in very wet areas making hydrochory a likely mechanism. Biotic vectors have also been observed. Scleria nutlets are often reported to be dispersed by birds (Bauters, 2018 and references therein). Other authors have observed ant-mediated dispersal (Gaddy, 1986), most commonly in species with tubercle-like structures near the base of the nutlet (e.g. subgenus Trachylomia). While dispersal via gravity and/or wind and ant-mediated dispersal likely occurs over short distances, dispersal via water and birds may happen over long distances (LDD).
Some species have local uses as medicines and materials (Simpson & Inglis, 2001), but they are not cultivated and have not purposefully been introduced outside their native distribution ranges. A few species have been reported as invasive (e.g. Scleria lacustris C.Wright in Florida; Jacono, 2001 (Moore & Donoghue, 2007)? In case of a shift in diversification rates, how often is it related to biogeographical movement (dispersification), trait change (key innovation), both or none? We hypothesize that shifts in diversification rates are significantly related with biogeographical movements and trait changes (niche shifts). In this study, we use the pantropical sedge genus Scleria to infer diversification rate patterns and their relationship with biogeographical movements (dispersification) and niche shifts (as a proxy of key innovation related to novel niche invasions).

| Phylogenetic analyses and divergence time estimation
ITS, ndhF and rps16 sequences were automatically aligned using Muscle (Edgar, 2004). Phylogenetic and divergence time estimations were performed in beast 2.4.5 (Bouckaert et al., 2014) using two GTR+I+G DNA substitution models for two independent partitions (nuclear versus. plastid DNA), a Birth-Death tree model, and an uncorrelated log-normal relaxed clock model (Drummond, Ho, Phillips, & Rambaut, 2006). We ran three independent analyses of 100 million generations. We evaluated mixing, convergence and stationary distribution using Tracer 1.7 (Rambaut, Drummond, Xie, Baele, & Suchard, 2018). Three calibration points, two secondary calibrations and one fossil calibration, were selected based on previous studies (Escudero & Hipp, 2013;Smith, Collinson, Rudall, & Simpson, 2010;Spalink et al., 2016). We calibrated the crown node of Scleria using the age of the oldest known fossil for the genus (Smith et al., 2010) with an offset of 33.8 Myr and standard deviation (SD) of 1.25 Myr. We applied a secondary calibration for the stem node of Scleria based on Spalink et al. (2016) with a mean of 57 Myr and SD of 2.5 Myr. Finally, we constrained the crown node of the sister group, tribe Bisboeckelereae, with a secondary calibration using a mean of 20 Myr and SD of 2.5 Myr (Escudero & Hipp, 2013). We applied a normal distribution prior to the secondary calibration points and a lognormal distribution prior to the fossil calibration point. Selected nucleotide substitution models were based on the maximum Akaike information criterion (AIC) weight resulting from the analysis of each DNA region in jModelTest 2.1.3 (Darriba, Taboada, Doallo, & Posada, 2012

| Biogeographical analyses
We inferred ancestral ranges using the R (R Core Team, 2019) package BioGeoBEARS (Matzke, 2013). We included seven areas in our analyses: South America, Central America, North America, Africa, Madagascar, Eurasia and Oceania (Table S2). We set these areas based on Dupin et al. (2017) with minor modifications considering peculiarities of our study group (specifically, we merged the Caribbean with Central America and we split Madagascar from Africa). The parameter maxareas was unconstrained. BioGeoBEARS implements two main models for large-scale biogeographical reconstruction: DIVA-like (Dispersal-Vicariance Analysis, Ronquist, 1997) and DEC (Dispersal-Extinction-Cladogenesis, Ree, Moore, Webb, & Donoghue, 2005;Ree & Smith, 2008). DEC and DIVA-like were also combined with the extra free parameter founder (j), which allows for cladogenetic dispersal, where the speciation event occurs in a different area than that of the ancestor. We also re-ran these four models incorporating into our models a matrix of dispersal connectivity corrected by a new free parameter w with temporal shifts in the potential dispersal con- given all eight models using AICc. We compare the fit of DEC models against DIVA-like models. We also compared the fit of models including the free parameter w and dispersal connectivity that change through time against models without those. We did not compare models with j and without j parameter because the way j parameter enters into the model does not allow such comparison (Ree & Sanmartín, 2018). The results obtained with models that include the parameter j will be interpreted cautiously as they tend to overvalue the role of cladogenetic dispersal (j) at the cost of underestimating (d). We performed biogeographical stochastic mapping (BSM) analyses as implemented in Matzke (2014) and Dupin et al. (2017) to estimate the number and type of biogeographical events. We conducted BSMs using the best-fitting models (DEC+w and DEC+j+w, see results). Event frequencies were estimated by taking the mean and SD of event counts from 50 BSMs.

| Diversification pattern analyses
Patterns and shifts in diversification rates were estimated in Bayesian analysis of macroevolutionary mixtures (BAMM) using reversible-jump Markov chain Monte Carlo (rjMCMC) (Rabosky, 2014).
The method allows changes in the numbers and locations of nodes at which speciation and extinction rates shift. All priors were set as recommended using the setBAMMpriors function (the analysis was conducted using a prior of one shift in diversification rates). The analysis was conducted assuming a global sampling fraction of 0.5 to account for missing taxa. The rjMCMC was run using Metropoliscoupling with four chains of five million generations each, saving trees every 1,000 generations. The R packages 'coda' (Plummer, Best, Cowles, & Vines, 2006) and BAMMtools (Rabosky et al., 2014) were used to check the Bayesian analysis and summarize and plot the results.

| Niche evolution analyses
To estimate the climatic niche of the taxa included in the phylogenetic analyses, a database of Scleria occurrences was built. All geo-   (2019). Three independent MCMC analyses of 2.5 million generations were run for each of the six bioclimatic variables.
We used a burn-in of 30%.
In order to study the evolution of the whole niche rather than single bioclimatic variables, we calculated principal components using the function prcomp (scale was set as true) implemented in R (R Core Team, 2019). We studied the evolution of PC1, PC2 and PC3 using bayou (with the same options as for the single bioclimatic variables).

| Quantitative state speciation and extinction
In order to model niche evolution and diversification rates, we used the model QuaSSE (FitzJohn, 2010) as implemented in diversitree (FitzJohn, 2012). We modelled the relationship between trait evolution and extinction rates as constant and the relationship between trait evolution and speciation as constant, linear, sigmoid and hump (FitzJohn, 2010 Model selection was performed using AIC.

| Phylogenetic analyses and divergence time estimation
Scleria split from its sister lineage c. 55.3 Ma (Table 1; Figure S1).

| Ancestral range estimations
To find the best model in BioGeoBEARS, we compared unconstrained versus constrained models, and DEC versus DIVA-like models. The fit (smaller AICc) was significantly better for constrained and DEC models, accordingly, we selected DEC+w and DEC+w+j as the best models (Table 2). The high dispersal rates inferred for Scleria make reconstruction of ranges for deep nodes of the phylogeny as highly equivocal (Figure 2; Figure S2). Main dispersal and coloniza-

| Diversification patterns
We found that five very similar scenarios accumulated 0.98 of posterior probability (PP, from the best to worst: 0.32, 0.29, 0.20, 0.12 and 0.057; Figure 5; Figure S4, Table S5

| Niche evolution
For all analyses, we obtained effective sizes for our Bayesian models >100. We report only the shifts with a PP >0.30. For details see Figure S5 and Table S6. Annual mean temperatures (bio1) showed low rates of stochastic evolution and high rates of evolution towards the optima. We inferred two shifts of optima for bio1 ( Figure 6a subgenus Scleria is from more temperate to a more tropical climatic regime, whereas that in subgenus Trachylomia is from a tropical to a more temperate climatic regime in temperature (lower temperature, higher range and seasonality) but the opposite in precipitation (less seasonality). In the sections of subgenus Scleria, the shifts are both to a more tropical climatic regime (sections Abortivae and Elatae) and to a more temperate climatic regime (core section Foveolidia).
Because Scleria species grow only in tropical areas (see Figure 1) the species and clades with a more temperate climatic regime are The results from the analyses of evolution of PC1, PC2 and PC3 ( Figure S6,   Figure S7). For the principal components, the models BM.contant.contant and OU.constant.constant are also significantly rejected. PC2 follows a BM process, whereas PC1 and PC3 follow an OU process. The principal components seem to better modelled by a hump.constant model in which speciation rates are highest at mid values (Table S8).

| Testing hypotheses of clade-specific diversification rates with BayesRate
Regarding hypotheses of diversification rates linked to niche shifts (Table S9)

| Are shifts in diversification rates linked to biogeographical movements?
We have inferred many biogeographical movements in Scleria then Scleria stands out even more as a high disperser. When we compare dispersal rates in Scleria with those in Cyperaceae overall (Spalink et al., 2016), dispersal rates are higher in Scleria. This is also true when we compare with other plant lineages, at genus (Chomicki & Renner, 2016;Echeverría-Londoño, Särkinen, Fenton, Purvis, & Knapp, 2020;Yao, Song, Yang, Tan, & Corlett, 2020)  Poaceae, a family with a similar graminoid habit, Hackel et al. (2018) found lower rates of dispersal in some tribes but higher rates of dispersal in others.

| Are shifts in diversification rates linked to major historical climatic events?
The  (Zachos et al., 2001). Shifts in diversification rates associated with major historical climatic events can entail F I G U R E 6 Niche evolution. (a) Phenotype bio1, (b) Phenotype bio7, (c) Phenotype bio18. Time in millions years from the origin of the genus Scleria to present is shown on the x axis. Evolution of the phenotype of bio1 (in °C*10), bio7 (in °C*10) and bio18 (in mm) is shown on the y axis. The evolution of phenotypes of species, nodes and branches across time and the phylogenetic relationships are plotted. Black branches indicate ancestral optimum and optima shifts are indicated with different colours. The name of each clade with an optimum shift is indicated with a colour matching with the colour of the branches [Colour figure can be viewed at wileyonlinelibrary.com] major extinctions in some clades, whereas diversification bursts can occur in others (Erwin, 2009

| Are dispersification events linked to niche evolution?
Our inferred niche shifts cannot be considered key innovations as none of the five inferred clades with niche shifts matches with the two inferred shifts in diversification rates. Nevertheless, QuaSSE analyses clearly support a relationship between the bioclimatic variables (and principal coordinates) and diversification rates. This means that although niche shifts do not provoke shifts in diversification rates, niche evolution is indeed shaping the diversification process in Scleria. This is also supported by our clade-specific diversification rate analyses, since, although the partition based on BAMM results is the most supported one, the partitions based on niche shifts were significantly supported in comparison with a single diversification rate regime.
Dispersification may couple synergistically with key innovations (Moore & Donoghue, 2007). Whereas the inferred niche shifts in Scleria (c. 30 Ma in subgenus Scleria towards a more tropical climatic regime, c. 20 Ma in subgenus Trachylomia towards a more temperate climatic regime and c. 7-9 Ma in several sections within subgenus Scleria towards both more temperate and more tropical climatic regimes) neither match major historical climatic events, nor inferred dispersification events. In this way, dispersification from South America to Africa without a climate niche shift seems to explain the shift in diversification rates in section Hypoporum suggesting that species were exapted. Shifts in climate niche evolution predate the second shift in diversification rates which suggest these were also exapted Nevertheless, within this clade (subgenus Scleria), the colonizations of Asia and Madagascar by sections Elatae and Abortivae, respectively, are coupled with two niche shifts suggesting that these colonizations involved the coetaneous climate niche adaptation of these clades but without subsequent shifts in diversification rates.

| Final remarks
We found high dispersal rates in Scleria, a genus with multiple dispersal syndromes. Shifts in diversification rates in Scleria are related either to biogeographical movement, or to both biogeographical movement and major historical climate events. However, shifts in diversification rates seem unrelated to niche transitions. Our results do not conclusively answer the question of why some biogeographical movements and/or trait changes implicate shifts in diversification rates, whereas others do not.

DATA AVA I L A B I L I T Y S TAT E M E N T
Sequence data are available from GenBank (see Table S1).
Distribution data were sourced from GBIF and georeferenced herbarium specimens available in accessible herbaria (see Material and Methods). The alignments and full set of occurrence data used in this study can be downloaded from DRYAD (https://doi.org/10.5061/ dryad.bnzs7 h486).