Being in the Right Place at the Right Time?

Home , Amsinckia calycina, Andersonglossum, Cynoglossum amabile, Diaspore (botany), Gyrocaryum, Heterocaryum

See discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/331592845

Being in the right place at the right time? Parallel diversiﬁcation bursts favored by the persistence of ancient epizoochorous traits and hidden factors in Cynoglossoideae

Article in American Journal of Botany · March 2019 DOI: 10.1002/ajb2.1251

CITATIONS READS 5 566

4 authors:

Ana Otero Pedro Jiménez-Mejías Spanish National Research Council Universidad Autónoma de Madrid

13 PUBLICATIONS 38 CITATIONS 142 PUBLICATIONS 1,496 CITATIONS

SEE PROFILE SEE PROFILE

Virginia Valcárcel Pablo Vargas Universidad Autónoma de Madrid Spanish National Research Council

62 PUBLICATIONS 711 CITATIONS 539 PUBLICATIONS 7,084 CITATIONS

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Evolutionary history of Antirrhinum genus View project

Completion of the Euro+Med Plantbase View project

All content following this page was uploaded by Pedro Jiménez-Mejías on 13 March 2019.

The user has requested enhancement of the downloaded file. RESEARCH ARTICLE

Being in the right place at the right time? Parallel diversification bursts favored by the persistence of ancient epizoochorous traits and hidden factors in Cynoglossoideae

Ana Otero1,2,4,5 , Pedro Jiménez-Mejías3,4, Virginia Valcárcel3,4, and Pablo Vargas1

Manuscript received 8 October 2018; revision accepted 14 January PREMISE OF THE STUDY: Long-distance dispersal (LDD) syndromes, especially endozoochory, 2019. facilitate plant colonization of new territories that trigger diversification. However, few 1 Departamento de Biodiversidad, Real Jardín Botánico, CSIC. Pza. de studies have analyzed how epizoochorous fruits influence both range distribution and Murillo, 2, 28014 Madrid, Spain diversification rates. We examined the evolutionary history of a hyperdiverse clade of 2 Escuela Internacional de Doctorado, Universidad Rey Juan Boraginaceae (subfamily Cynoglossoideae, eight tribes, ~60 genera, ~1100 species) and Carlos, C/Tulipán s/n, 28933 Móstoles, Spain the evolution of fruit traits. We evaluated the evolutionary history of diaspore syndromes 3 Centro de Investigación en Biodiversidad y Cambio Global (CIBC- UAM), Universidad Autónoma de Madrid, 28049 Madrid, Spain correlated with geographic distribution and diversification rates over time. 4 Departamento de Biología (Botánica), Facultad de Ciencias METHODS: Plastid DNA regions and morphological traits associated with dispersal Biológicas, Universidad Autónoma de Madrid, C/ Darwin, 2, 28049 syndromes were analyzed for 71 genera (226 species). We employed trait-dependent Madrid, Spain diversification analysis (HiSSE) and biogeographic reconstruction (Lagrange) using a 5 Authors for correspondence (e-mail: [email protected], [email protected]) time-calibrated phylogeny. Citation: Otero, A., P. Jiménez-Mejías, V. Valcárcel, and P. Vargas. KEY RESULTS: Our results indicate that (1) the earliest divergence events in Cynoglossoideae 2019. Being in the right place at the right time? Parallel diversification occurred in the central-northeastern Palearctic during the Paleogene (early to middle bursts favored by the persistence of ancient epizoochorous traits Eocene); (2) an epizoochorous trait (specialized hooks named glochids) is ancestral and has and hidden factors in Cynoglossoideae. American Journal of Botany 106(3): 1–15. been maintained long term; and (3) glochids are correlated with increased diversification doi:10.1002/ajb2.1251 rates in two distantly related clades (Rochelieae and Cynoglossinae). Rapid speciation occurred for these two groups in the same area (central-eastern Palearctic) and same period (Oligocene-Miocene: Rochelieae, 30.82–13.69 mya; Cynoglossinae, 33.10–15.21 mya). Lower diversification rates were inferred for the remaining four glochid-bearing clades.

CONCLUSIONS: One more example of “biogeographic congruence” in angiosperms is supported by a shared geographic (central-northeastern Palearctic) and temporal (28.60–21.59 mya, late Oligocene) opportunity window for two main clades’ diversification. Epizoochorous traits (fruit glochids) had an effect in higher diversification rates only with the joint effect of other unmeasured factors.

KEY WORDS biogeography; Boraginaceae; diversification rates; HiSSE; hyperdiverse; phylogeny; trait-dependent reconstruction; historical contingency.

Fleshy fruit innovations appearing in late Cretaceous angiosperms increases the chances of establishing new populations and con- have long been associated with the diversification of vertebrates sequently promoting speciation (see Beaulieu and Donoghue, that acted as seed dispersers (Regal, 1977). Indeed, mammal and 2013). Likewise, the acquisition of a trait related to biotic disper- bird diversification during the Cenozoic would have favored the sal is a key innovation that may have boosted diversification rates evolution of fruits specialized for both internal (endozoochory) in many taxonomic groups (e.g., endozoochory in Myrtaceae and external (epizoochory) dispersal by vertebrates (Burger, [Biffin et al., 2010] and endozoochory and epizoochory in Fagales 1981; Tiffney, 1986, 2004). Biotic dispersal also favors geographic [Larson-Johnson, 2016]). However, while the acquisition or loss range expansion through the colonization of new areas, which of a single trait may promote diversification, the cumulative action

of various factors must be invoked in the spatiotemporal context 2014). Interestingly, even though Cynoglossoideae is the most of each particular lineage (e.g., see “synnovation” in Donoghue diversified subfamily, a notable unevenness of species diversity and Sanderson, 2015). exists among the lineages. There are many rare, endemic genera Long-distance dispersal (LDD) has many biogeographic im- (e.g., Gyrocaryum, Myosotidium, Suchtelenia, Antiotrema) that plications; it explains the movement of diaspores over dispersal contrast with highly diversified and widely distributed lineages barriers to distant territories, between continent and islands, (e.g., Lappula, Cynoglossum, Myosotis, Mertensia) (Chacón et al., and within large continents. Previous studies indicated that LDD 2016). through epizoochory is relatively rare (<5% of all angiosperm Integrative analyses of historical biogeography and diversi- species studied in 10 regional floras [Sorensen, 1986]; 5.2% for the fication rates are needed to properly evaluate whether diaspore entire flora of Europe [Heleno and Vargas, 2015]), and that epizo- specializations related to LDD have been driving factors in mac- ochorous traits are less effective than endozoochorous traits, es- roevolutionary patterns of speciation, extinction, and dispersal pecially with avian vectors (Costa et al., 2014; Heleno and Vargas, (Beaulieu and Donoghue, 2013; Larson-Johnson, 2016). The high 2015). Despite low occurrence in angiosperms, epizoochorous number of widespread clades and lineages of Cynoglossoideae that traits have been invoked to explain species connections via in- occur in more than three continents (five lineages:Trichodesma , tercontinental (Vargas et al., 1999) and continent-island dispersal Myosotis, Lappula, Mertensia, and Hackelia; two clades: subtribe (Heleno and Vargas, 2015). Indeed, Sorensen (1986) proposed Cynoglossinae and tribe Omphalodeae) and the predominance of that epizoochory is able to successfully disperse diaspores over fruit specializations related to LDD syndromes (Otero et al., 2014) larger geographic scales than all other LDD syndromes (i.e., fruit offer a unique opportunity to assess the role of diaspore traits in specializations related to LDD: anemochory, thalassochory, en- historical biogeography and diversification. dozoochory). However, no studies have addressed whether high Considering the worldwide distribution of Cynoglossoideae, diversification rates exhibited by widespread plant groups were its diaspore diversity (unique in the family), and the uneven initially promoted by the acquisition of certain diaspore traits. To diversification within clades, we hypothesize that LDD fruit better elucidate this relationship, new trait-dependent diversifi- specialization resulted in geographic range expansion and an in- cation analyses have been developed. Recently, Larson-Johnson creased rate of diversification. To test this hypothesis, we recon- (2016) has suggested that biotically dispersed lineages of Fagales structed the biogeographic history and morphological evolution display higher diversification rates than abiotically dispersed of fruits across a time-calibrated phylogeny of Cynoglossoideae. ones. Nevertheless, such investigations are still needed for other Specifically, we (1) characterized diaspore functional traits asso- hyperdiverse angiosperm groups. ciated with LDD syndromes using SEM, light microscopy, and The origin of Boraginaceae (105 genera, ~1600 species) has literature sources; (2) reconstructed the evolution of diaspore been dated to the late Cretaceous in Eurasia (Luebert et al., 2017). (typically fruits) syndromes across the phylogeny; and (3) eval- This family is characterized by schizocarpic fruits that split into uated whether diaspore traits related to LDD (epizoochorous) four one-seeded units (although some exceptional species have favored large geographic distributions and high diversification one or two units; “nutlets,” Weigend et al., 2013). The most recent rates. phylogenetic studies identify three subfamilies that differ not only in number of species but also in their distribution ranges: Boraginoideae (~610 spp., primarily in the western Palearctic MATERIALS AND METHODS and the Americas), Echiochiloideae (~30 spp., in North America, Africa, and southwestern Europe), and Cynoglossoideae (~1100 Taxon sampling spp., cosmopolitan, widespread on all continents except the Antarctic) (Chacón et al., 2016, 2017). These subfamilies are A total of 47 genera and 192 species of the ~50 genera and ~1100 characterized by different fruit types (Gürke, 1893; Hilger, species recognized in Cynoglossoideae (Chacón et al., 2016; 2014). While the nutlets in Boraginoideae and Echiochiloideae Appendix S1) were sampled. Therefore, the diversity of the sub- are glabrous and slightly wrinkled or smooth, Cynoglossoideae family is well represented, with 94% of the genera sampled de- have the largest diversity of nutlet shape and ornamentation. spite the relatively low species sample size (18%) (Appendix S2). These fruits range from smooth to densely covered with hook- Between one and 23 species of each lineage at genus level (ge- like glochids, and from wingless to winged nutlets (Otero neric lineage) were sampled in order to represent the geographic et al., 2014). Traditionally, taxonomic classifications of the range and morphological variation within each of these lin- Cynoglossoideae were based on nutlet shape and ornamentation eages. Therefore, all morphological variation of fruits, and thus (Gürke, 1893; Popov, 1953; Zhu et al., 1995). However, recent all dispersal syndromes of Cynoglossoideae, was included. Two phylogenetic studies revealed that many of these character states monospecific genera (Antiotrema and Stephanocaryum) and 48 are not synapomorphies and that the taxonomic groups they de- other species (a total of 27 genera) were sequenced for the first limit are artificial (Selvi et al., 2011; Weigend et al., 2013). The time. The remaining 144 species (20 genera) of Cynoglossoideae great morphological variation among fruits of Cynoglossoideae were obtained from GenBank (NCBI, Bethesda, Maryland, USA) has been associated with different dispersal syndromes, mainly (Appendix S1). Additionally, sequences of 24 genera (34 species) epizoochory and anemochory (Cohen, 2013). In particular, epi- representing other subfamilies in Boraginaceae and families in zoochorous species have been associated with the development Boraginales were obtained from the GenBank and used as out- of various hooked trichomes, glochids, and bristles on fruit and group taxa (Appendix S1). In total, our study included 71 genera calyx for attaching to animal fur and plumage (Selvi et al., 2011; (226 species) in Boraginales. Hilger, 2014). Otherwise, anemochorous Cynoglossoideae pro- The specimens were obtained from herbarium specimens reduce nutlets with broadly winged rims (Hilger, 2014; Otero et al., ceived on loans from 16 herbaria, as well as our own field collections 2019, Volume 106 • Otero et al.—Epizoochory coupled with lurking factors boosted diversification • 3

(Appendix S1). Taxonomic literature was used to identify the spec- the evolution of the characters that allowed the assignment may imens (Appendix S2). have appeared somewhere between stem and crown node of a clade (Renner, 2005). In particular, we set the values for lognor- DNA extraction and sequence matrices mal mean (logMean), lognormal standard deviation (logSD), and offset according to previous studies (Chacón et al., 2017; Luebert DNA was extracted from leaf tissue using DNeasy Plant Mini et al., 2017) and used the following calibration points: (1) clade of Kits (Qiagen, Valencia, California, USA). The plastid genome has Ehretia (lognormal prior, logMean = 1.4, logSD = 1.0 offset = 47.8 been widely used in plant phylogenetics because of its haploid and mya; Chandler, 1964); (2) clade of Tournefortia (lognormal prior, non-recombinant nature. When the plastome is maternally inher- logMean = 1.1, logSD = 1.0 offset = 28.1 mya; Graham and Jarzen, ited (i.e., transmitted via seeds as in Boraginaceae; Corriveau and 1969); (3) clade Echiochiloideae (lognormal prior, logMean = Coleman, 1988), plastid sequences are suitable for inferring pat- 2.0, logSD = 1.0 offset = 41.2 mya; Hammouda et al., 2015); (4) terns of colonization by seed dispersal (Avise and Riddle, 2009). clade of Lithospermum (lognormal prior, logMean = 0.5, logSD = Moreover, the framework of molecular data in Boraginaceae is 1.0 offset = 10.3 mya; Gabel, 1987); (5) clade of Lappula and primarily plastid based, and plastid loci are used to maximize Rochelia (lognormal prior, logMean = 0.5, logSD = 1.0 offset = species representation. Therefore, two DNA plastid regions were 10.3 mya; Gabel et al., 1998); (6) clade Amsinckiinae (excluding chosen on the basis of previous studies of Boraginaceae (Weigend Andersoglossum and Adelinia) (lognormal prior, logMean = 0.5, et al., 2013; Chacón et al., 2017): the trnL-trnF intergenic spacer logSD = 1.0 offset = 10.3 mya; Segal, 1966; Gabel et al., 1998). (including trnL intron) and the rps16 intron. Primers for the trnL- Convergence and effective sample size (ESS) values for all pa- trnF region were used as in Otero et al. (2014). For the rps16 re- rameters were assessed using Tracer version 1.6 (Rambaut et al., gion, we used the two external primers rps16 F and rps16 R2 2014). Subsequently, LogCombiner version 1.8.2 (Drummond (Oxelman et al., 1997) plus two newly designed internal prim- et al., 2012) was used to combine the eight runs. Resampling was ers: rps16 F10 (5′ GCTCYTYRYTCGACATC 3′) and rps16 R766 done by selecting one of every two trees from the 10,000 trees (5′ TACTCATAACTCAAGTTG 3′). Polymerase chain reaction obtained per run, to reduce the high number of trees obtained. conditions described by Otero et al. (2014) were used for both DNA Finally, TreeAnnotator version 1.8.2. (Drummond et al., 2012) regions, and the resulting products were sequenced by Macrogen was run to obtain the maximum clade credibility (MCC) tree Europe (Amsterdam, The Netherlands). A concatenated matrix of by reconstructing the mean node height after a burn-in of 25% the two DNA regions for 226 species was automatically aligned of trees. To assess the robustness of the phylogeny recovered using MAFFT (similarity-based alignment; Katoh and Standley, in BEAST, Bayesian inference and maximum likelihood anal- 2013). After visualizing the alignment in Geneious version 9.0.2 yses were run in MrBayes version 3.2.6 (Ronquist et al., 2012) (Kearse et al., 2012), we performed minor adjustments to mini- and RAxML version 8.2.10 (Stamatakis, 2014), respectively. For mize ambiguous positions at flanking gaps derived from the auto- Bayesian inference, we set two runs of 50 million generations, mated alignment. Data for rps16 were missing for only six samples a general time-reversible (GTR) model for both trnL-trnF and (Appendix S1). The nucleotide substitution model that best fit each rps16, and 25% burn-in. Values of ESS were reviewed in Tracer partition (trnL-trnF and rps16) was inferred using the Akaike in- to determine the convergence of the two runs. For maximum formation criterion corrected for small sample size (AICc) through likelihood analysis, we also used a GTR model for each parti- jModelTest version 0.1.1 (Posada, 2008). tion (trnL-trnF and rps16) and rapid bootstrap searching with extended majority-rule consensus (AutoMRE) as bootstrap stop- Estimation of divergence times and phylogenetic ping criterion. reconstructions Biogeographic analyses Divergence time reconstructions were performed under Bayesian inference using BEAST version 1.8.2 (Drummond et al., 2012) Biogeographic reconstructions were performed using dispersal- on the CIPRES portal (Miller et al., 2010). We conducted eight extinction-cladogenesis (DEC) analysis as implemented in independent Markov chain Monte Carlo runs of 100 × 106 gen- Lagrange version 2.0.1 (Ree and Smith, 2008), using the MCC tree erations, sampling every 10,000 generations, and each with four as input, to determine the most probable ancestral distribution. chains. Uncorrelated lognormal relaxed clock and a birth-death Both the times of diversification posterior to tectonics plate move- tree model were selected for the two partitions of the plastid ma- ments and the great colonization ability for most generic lineages, trix (trnL-trnF and rps16). A uniform distribution prior with val- which are currently present in more than three continents, led us to ues between 1.0E-4 and 1.0E-2 was set for the relaxed clock in select DEC. This model is better than other models such as DIVA, order to facilitate the convergence of Markov chains and improve in which times of diversification cannot be included (vicariance is analysis efficiency. The chosen boundaries followed Blanco- favored respect to dispersal) and widespread inheritance of more Pastor et al. (2012). Separate nucleotide substitution models were than two areas by only one of the two descendants is not permit- considered for each of the two plastid regions. Seven points of ted (I. Sanmartín, personal communication). Given that evolution- calibration were used to date divergence times: (1) the root, set to ary (nonstatistical) criteria for model choice is more appropriate 61.18 mya based on Magallón et al. (2015) (crown of Boraginales: than inadequate model choice based on AIC (Ree and Sanmartín, normal prior, mean = 61.18 mya, SD = 12); and (2) six fossils that 2018), we selected the DEC model as the most suitable for our study have been confidently assigned to phylogenetic nodes and tested case. Seven geographic areas were delimited in order to represent in previous studies (Luebert et al., 2011, 2017). The calibration the spatial distribution and areas of endemicity of the generic lin- points were set in all cases to the stem node. This is a conserva- eages of Cynoglossoideae: (A) western Palearctic (W Palearctic: tive approach for fossils that can be assigned to extant taxa, since Europe, Mediterranean Africa, and southwestern Asia), (B) North 4 • American Journal of Botany

America, (C) South America, (D) central-northeastern Palearctic In order to identify diaspore traits associated with different (C-NE Palearctic, including the Himalayas), (E) southeastern syndromes, we characterized nutlets for 36 genera (63 species) Asia (SE Asia: India, Indochina, eastern China, and Japan), (F) and 32 genera (44 species) using scanning electronic microscopy Sub-Saharan Africa, and (G) Australasia-Oceania (including New (SEM, Hitachi S3000N) and light microscopy (LM; Olympus BX51) Zealand). These seven areas were based on Chacón et al. (2017) (Appendix S3) at Real Jardín Botánico and Museo de Ciencias and Luebert et al. (2017), with special emphasis on the main areas Naturales (Madrid, Spain). In addition, nutlet morphology of five of endemicity observed across lineages (e.g., existence of endemic genera (six species) from the sister subfamily Boraginoideae was genera in SE Asia: Antiotrema, Nihon, Sinojohnstonia, Thyrocarpus, also studied under SEM/LM (Appendix S3). The SEM screening and Trigonotis; Zhu et al., 1995; Kadota, 2009). These areas were procedure developed by Otero et al. (2014) was used. Cross sections assigned to species according to their current distributions. Each of resin-embedded nutlets were stained with 2% toluidine blue and generic lineage distribution was well represented, given that the imaged at the Microscopy Service of University of Seville (CITIUS, species included in the phylogeny represent the whole range of Seville, Spain). Digital images from online repositories (Simpson, each genus. No temporal limitations were specified for dispersal; 2012; Plants, 2014) were used for 131 species (Appendix S2; se- all seven areas were allowed to be ancestral for any node, with up quences are deposited in GenBank), but no fruits were available in to four areas as maximum range size. Adjacency was allowed where the herbarium specimens examined. extant ranges share borders or those supported by the fossil record and previous molecular investigations (Chacón et al., 2017; Luebert Evolution of diaspore traits et al., 2017). In particular, adjacency was allowed between North and South America, W Palearctic and C-NE Palearctic, C-NE Palearctic The diaspore trait characterization combined with bibliographic and SE Asia, North America and C-NE Palearctic (Donoghue and metanalysis (Appendices S2, S3) let us identify six main diaspore Smith, 2004), South America and Oceania (Thorne, 1972), and SE characters (five binary and one multistate): (1) glochids (0 = ab- Asia and Oceania (Raven, 1972). Sub-Saharan Africa was consid- sence, 1 = presence); (2) hooked trichomes on nutlet (0 = absence, ered adjacent to W Palearctic. In order to obtain a comprehensible 1 = presence); (3) hooked trichomes on calyx (0 = absence, 1 = pres- visualization of Lagrange reconstruction, relative probabilities of ence); (4) nutlet margin shape (0 = absence, 1 = wide and open, 2 = each area combination (independently of the split frequency) were wide and incurved, 3 = narrow); (5) deeply dentate nutlet edge (0 = summarized and represented in pie format for each node (see be- absence, 1 = presence); and (6) mesocarpic protrusion (0 = absence, low). The area assignment for the outgroup species was done as ex- 1 = presence). Three dispersal syndromes were recognized, based plained above, although for the sake of simplicity the ancestral areas on specific sets of diaspore characters: (1) epizoochorous traits reconstructed for the outgroup species are not shown in Figure 2. (dispersal externally by animals), herein identified as presence of hooked trichomes, glochids or dentate margins (0 = absence, 1 = Diaspore trait and syndrome characterization presence); (2) anemochorous traits (those that can promote dispersal by wind), herein identified as air flotation assisted by a nutlet The multiple dispersal syndromes in Cynoglossoideae are correlated margin wide and open (0 = absence, 1 = presence); and (3) myr- with diverse fruit morphology (Cohen, 2013). Epizoochorous and mecochorous traits (those that are associated with dispersal by anemochorous traits are predominant in Cynoglossoideae. The ants), herein identified by a mesocarpic protrusion (0 = absence, most prominent and complex epizoochorous structures are special- 1 = presence). The endozoochorous traits of outgroup species (i.e., ized fruit structures named glochids. These structures are barbed fleshy fruits dispersed via animal ingestion) were treated as a fourth hooks reinforced by silicon or calcium; they only occur in this sub- dispersal syndrome. family within the order Boraginales. In particular, glochids broadly Ancestral character-state reconstruction was performed for characterize the tribe Rochelieae and the subtribe Cynoglossinae both diaspore characters and dispersal syndrome (all except endo- of tribe Cynoglosseae (Selvi et al., 2011; Hilger, 2014). Additional zoochorous traits). In addition, a general ancestral character-state adherent structures may also be found on nutlets—that is, hooked reconstruction including the five syndrome categories was done to hairs in genus Iberodes (Otero et al., 2014), deeply dentate margins summarize the results (0 = unspecialized, 1 = epizoochorous, 2 = in Mimophytum (Holstein et al., 2016)—as considered in previous anemochorous, 3 = myrmecochorous, 4 = endozoochorous). Eight studies (Selvi et al., 2011; Hilger, 2014; Fig. 1). Additional structures of the 192 species have polychorous traits (i.e., set of traits related on the sepals (i.e., bristles and hooked hairs in Amsinckiinae and to more than a single dispersal syndrome). But since their diaspore Myosotideae) allow the entire schizocarp to disperse as a unit inside glochids were more apparent than their other traits, these species the calyx (Hasenstab-Lehman and Simpson, 2012). Anemochorous were scored as epizoochorous. A stochastic mapping approach was traits in Cynoglossoideae are characterized by the development of conducted using the function “make.simmap” in the R package an open wide wing from the nutlet rim (Hilger, 2014; Otero et al., “phytools” (R Core Team, 2012; Revell, 2012) allowing different 2014; Fig. 1). Besides LDD by epizoochory and anemochory, there transition rates between all states. The MCC tree was the input tree is poor information about other types of short-distance dispersal for stochastic mapping. After simulating 200 stochastic histories, a syndromes documented in Cynoglossoideae, such as myrmeco- summary tree was obtained from each character reconstruction. chory. In our study, we considered traits related to myrmecochory when the mesocarp protrudes to the exterior of the nutlet and Trait-dependent diversification rates and epizoochory forms an excrescence either on the abaxial or the adaxial nutlet side (Fig. 1K, M) (see Gorb et al., 2000; Gorb and Gorb, 2003). Finally, Diversification rates associated with presence–absence of epi- some species display more than one dispersal syndrome on the di- zoochorous traits were examined in Cynoglossoideae. Hidden aspore (diplochory), combining epizoochory, myrmecochory, or state speciation and extinction (HiSSE; Beaulieu and O’Meara, anemochory (Burkart, 2000; Melcher et al., 2000). 2016) analysis was used to test whether glochids were correlated 2019, Volume 106 • Otero et al.—Epizoochory coupled with lurking factors boosted diversification • 5

1 mm A B 1 mm C D

1 mm 1 mm

350 µm 2 mm E F 50 µm G H

1 mm

1 mm 500 µm I J K L

500 µm 1 mm

M N 1 mm O P

50 µm 1 mm 50 µm

FIGURE 1. SEM photographs of the nutlets of some species of Cynoglossoideae representing epizoochorous and anemochorous diaspore traits and light microscopy stained microtome cross sections of the distinct mesocarpic emergence types. White arrows from A to E indicate the direction from proximal to distal extremes of nutlets. Black and white arrows from I to P indicate the point of mesocarpic emergence. (A–C) Epizoochorous traits. (A) Abaxial side of Lappula squarrosa (glochids). (B) Abaxial side of Iberodes littoralis (uncinated trichomes). (C) Abaxial side of Mimophytum alienoides (deeply dentate margin). (D, E) Anemochorous-related traits. (D) Abaxial side of the widely winged anemochorous Mimophytum alienum. (E) Abaxial side of the narrowly winged not-anemochorous Iberodes commutata. (F–H) Detail of distinct diaspore trait adaptations. (F) Glochids of L. squarrosa. (G) Uncinated trichomes of I. littoralis. (H) Wing surface of Rindera tetraspis. (I–P) Nutlet cross section of distinct types of mesocarpic emergences. (I, J) Abaxial ontogenetic emergence in Microula trichocarpa; mesocarpic emergence is located at the internal side of nutlet margin. (K, L) Abaxial ring-like emergence in Gyrocaryum oppositifolium. Ring-like mesocarpic emergence located at the internal side of nutlet margin. (M, N) Adaxial mesocarpic emergence in two species of Anchusa through the attachment scar. (M) Adaxial mesocarpic emergence in A. officinalis. (N) Adaxial mesocarpic emergence in A. undulata. (O, P) Adaxial mesocarpic groove in two species of Amsinckiinae. (O) Adaxial groove in Oreocarya johnstonii. (P) Adaxial groove of Cryptantha decipiens. to higher diversification rates. HiSSE allows inferring correlation rate (Beaulieu and O’Meara, 2016). The observation of hidden between diversification rates and an observed character (state 0, effects allows for the inclusion of unmeasured factors (e.g., life glochids absent; state 1, glochids present) when “hidden effects” history, biogeographic movements, physical traits) that are in- (unmeasured characters) are involved (state A, hidden effect fluencing diversification rate (Beaulieu and O’Meara, 2016). absent; state B, hidden effect present). Correlation between an Nevertheless, identification of these hidden factors is not the aim observed character and high diversification rates in the presence of this analysis, and in most cases they can be difficult to assess of a hidden effect suggests that multiple factors are acting to- confidently (Beaulieu and O’Meara, 2016). In addition, HiSSE ad- gether with the observed trait to trigger a higher diversification dresses some inherent limitations of previous state speciation and 6 • American Journal of Botany

FIGURE 2. Time-calibrated phylogeny and biogeographic reconstruction of Cynoglossoideae with names of the main taxonomic clades. Circles indicate the Bayesian posterior probability (BPP) for each node with values ≥0.84. White circles represent nodes with BPP values from 0.84 to 0.89. Gray circles represent nodes with BPP values from 0.90 to 0.94. Black circles represent nodes with BPP values from 0.95 to 1. Nodes numbered 1–6 indicate fossil calibration points as described in the text (1: Ehretia; 2: Tournefortia; 3: Echiochiloideae; 4: Lithospermum; 5: Lappula-Rochelia; 6: Amsinckiinae). The 95% highest posterior density (HPD) bars are provided for the crown age of main taxonomic clades (see Appendix S13 for HPD bar of remaining lineages). Relative probability of each area combination is represented in the pie chart of each node. Color legend and a map for each area are provided. The international stratigraphic scale is included from 73 mya until present. extinction models (BiSSE, MuSSE, GeoSSE) where the consider- (CID-2 and CID-4) that assume the evolution of a binary charac- ation of a simplistic null hypothesis often implied its rejection, ter independent from diversification process without forcing the resulting in false-positive correlations (Laenen et al., 2016). To diversification to be constant across the entire tree (Beaulieu and ameliorate this issue, HiSSE implements two types of null models O’Meara, 2016). We tested 44 models through the function ‘hisse’ 2019, Volume 106 • Otero et al.—Epizoochory coupled with lurking factors boosted diversification • 7

implemented in the R package “hisse” (Beaulieu, 2017; R Core (Rabosky, 2014) were performed (Appendix S5). Controversy about Team, 2012) to address more complex evolutionary scenarios the inferences yielded by BAMM analyses has been predominant from the null models, in agreement with the proposal of Laenen in recent years, with contrasting opinions (e.g., Moore et al., 2016; et al. (2016). The models differ in whether turnover rates (τ), ex- Rabosky et al., 2017; Meyer and Wiens, 2018; Rabosky, 2018). As a tinction fraction (ε), or the combination thereof are allowed to result, we considered BAMM results as complementary to HiSSE vary between states 0 and 1 and also whether the transition rates inferences. are free to vary among states (Appendix S4; Laenen et al., 2016). Finally, transition rates were allowed to vary between states of the observed character (glochidiate/non-glochidiate fruits) except RESULTS for those that also imply a change of the hidden state (e.g., 1A to 0B and 0A to 1B), since they are highly improbable according to Biogeographic and divergence time analyses HiSSE vignette (Beaulieu, 2016). In order to allow uncertainty in model selection, we plotted ancestral state reconstruction and in- The concatenated sequence matrix oftrnL-trnF and rps16 was ferred diversification rates by model averaging across all 44 mod- 2320 bp containing 855 parsimony-informative sites. The GTR els using Akaike weights through the function “plot.hisse.states.” model with gamma distribution of rates across sites was selected In addition, we represented averaged net diversification inferred as the best-fitting substitution model for both plastid regions, across all nodes and tips for each observed state (glochidiate/ trnL-trnF and rps16. The Bayesian and maximum likelihood re- non-glochidiate) through bean plots using the R package “bean- constructions obtained with MrBayes and RAxML were congru- lot” (R Core Team, 2012; Kampstra and Kampstra, 2014). ent with the one obtained from BEAST (Appendix S6). Subfamily To avoid the influence of diversification rates in genera out of Cynoglossoideae is divided into nine main clades (Fig. 2) that Cynoglossoideae, the outgroup was removed from the MCC tree correspond to the currently recognized tribes and subtribes to conform the input tree with the R package “ape” (Paradis et al., (Chacón et al., 2016). Tribe Trichodesmeae is inferred to be sister 2004; R Core Team, 2012). The assignment of the observed char- to a clade comprising all other clades of the subfamily (Fig. 2). acter state to the tips of a given generic lineage was based on the This clade is divided into two subclades: (1) a subclade that con- character state of the species included in the phylogeny. However, tains the tribes Asperugeae, Omphalodeae, and Rochelieae; and a correction was applied to maintain the proportion of glochidiate/ (2) a subclade containing tribes Myosotideae and Cynoglosseae. non-glochidiate species of the generic lineage when all the species Cynoglosseae are in turn divided into four subtribes: Microulinae were considered. Given the taxonomic importance of glochids, this and Amsinckiinae sister to Bothriosperminae and Cynoglossinae information was extracted from the literature (Appendix S2). (Fig. 2). Ancestral areas and mean divergence times inferred for Additionally, to explore the evolution of diversification rates generic lineages of Cynoglossoideae are shown in Fig. 2 and sum- within different clades of Cynoglossoideae separately, Bayesian anal- marized in Table 1. The ESS values of the divergence time estima- ysis of clade-specific speciation and extinction rates through BAMM tion parameters were >200, indicating the convergence of Markov

TABLE 1. Summary of the three main macroevolutionary indicators of main clades of Cynoglossoideae. The second and third columns show the mean age (myr) of each clade and highest posterior density (HPD) ranges for stem and crown nodes, respectively; the fourth column shows the ancestral area inferred in biogeographic reconstruction from the seven geographic areas delimited (A: western Palearctic; B: North America; C: South America; D: central-northeastern Palearctic; E: southeastern Asia; F: tropical Africa; and G: Australasia-Oceania); the fifth column shows the mean of diversification rate and standard deviation obtained in BAMM reconstruction. Dash (–) indicates nodes with Bayesian posterior probability <0.84. Mean age Mean age (95% HPD) (95% HPD) Ancestral area (relative Diversification rate Clades Stem node Crown node probability) (mean ± SD) Cynoglossoideae 49.25 46.04 D (0.89) 0.31 ± 0.04 (37.94–65.29) (33.42–64.10) Trichodesmeae 46.04 34.59 D (0.81) 0.15 ± 0.03 (33.42–64.10) (16.47–53.85) Omphalodeae – 16.59 ABCD (0.24), AD (0.19), ABD (0.21), 0.15 ± 0.02 (8.65–25.45) ABDE (0.12), A (0.07), ADE (0.03) Asperugeae 33.05 25.17 D (0.44), A (0.21), AD (0.13) 0.15 ± 0.03 (21.57–46.95) (13.72–37.62) Rochelieae – 21.59 D (0.79) 0.53 ± 0.13 (13.69–30.82) Myosotideae 38.14 29.72 D (0.2), DE (0.24) 0.18 ± 0.03 (27.03–53.18) (18.99–42.21) Microulinae 27.99 24.57 AD (0.19), DE (0.29), ADF (0.24), D 0.15 ± 0.02 (19.10–38.44) (15.94–34.31) (0.11) Amsinckiinae 27.99 22.11 B (0.93) 0.27 ± 0.09 (19.10–38.44) (14.33–31.45) Bothriosperminae 28.6 15.74 DE (0.79), D (0.10) 0.15 ± 0.02 (19.50–39.94) (7.20–24.94) Cynoglossinae 28.6 23.59 D (0.98) 0.64 ± 0.19 (19.50–39.94) (15.21–33.10) 8 • American Journal of Botany

chains. Subfamily Cynoglossoideae was inferred to have first di- and Cynoglossinae (C-NE Palearctic). During the early Miocene, versified during the mid-Eocene (mean crown age = 46.04 myr; the remaining two main clades, Omphalodeae (equivocal in the 95% highest probability density [HPD] = 33.42–64.09 myr) in C- Palearctic; Table 1) and Bothriosperminae, (widespread in C-NE NE Palearctic (relative probability = 0.90). The estimated crown Palearctic and SE Asia; Table 1) most likely diversified. age of Trichodesmeae coincides with the Eocene-Oligocene The earliest events of cladogenesis in most clades occurred boundary in the C-NE Palearctic (relative probability = 0.81). within the same ancestral area where they originated, such as in Crown age of Microulinae (Palearctic; Table 1) and Asperugeae Trichodesmeae, Rochelieae, Cynoglossinae (C-NE Palearctic from and Myosotideae (C-NE Palearctic) was inferred during the the late Eocene to the Miocene), Amsinckiinae (North America Oligocene. The Oligocene-Miocene boundary was inferred to be from the early to the middle Miocene), and Bothriosperminae the crown age of Amsinckiinae (North America) and Rochelieae (East Palearctic from middle to late Miocene) (Fig. 3; Appendix

Mid Eocene (46.04-33.90 mya) Oligocene (33.90-23.03 mya) Amsinckiinae (27-22 mya)

Early Miocene (23.03-13.82 mya) Late Miocene (13.82-5.33 mya) Rindera lanata-Pardoglossum (18-10 mya) Memoremea - Stephanocaryum (25-11 mya) Hackelia- Embadium- Asperugo-Mertensia (25-17 mya) Austrocynoglossum (15-8 mya) Myosotis (18-14 mya) Mertensia (17-9 mya)

Plagiobothrys Nihon (15-6 mya) (14-8 mya ) Trigonotis - Microula oblongifolia - Trichodesma Sinojohnstonia (29-16 mya) Pectocarya M.sikkimensis (13-6 mya) (10-8 mya) Bothriosperminae (28-15 mya) (13-5 mya) Microula turbinata - M.muliensis (9-4 mya) Microula myosotidea - M.ovalifolia (9-2 mya)

Pliocene (5.33-2.60 mya) Pleistocene-Holocene (2.60-Present) Asperugeae Rochelia (4-2 mya) Asperugeae Asperugeae Bothriosperminae Amsinckiinae Cynoglossinae Cynoglossinae Myosotideae Microulinae Myosotis incrassata- Myosotideae Omphalodeae Myosotideae M.macrosperma (10-5 mya) Cynoglossum amabile - Omphalodeae Trichodesmeae Rochelieae Trichodesmeae C.australe (18-3 mya) Rochelieae Rochelieae Bothriosperminae Cynoglossinae Cynoglossinae Amsinckia calycina - Amsinckiinae Microulinae Microulinae Myosotideae A.menziesii (8-2 mya) Omphalodeae Trichodesmeae Rochelieae Trichodesmeae Cynoglossinae Myosotideae* Omphalodeae Rochelieae Trichodesmeae

Epizoochorous traits Myrmecochorous traits Unspecialized

FIGURE 3. Main dispersal events detected in biogeographic analyses of Cynoglossoideae evolution. The evolutionary history of this subfamily has been divided in six geologic epochs indicated in millions of years ago (mya). Since dispersal events may occur at any point along branches between stem and crown nodes, the average between these two points was used as the dispersal event age. Nonetheless, stem and crown ages are provided in brackets for each dispersal event. Arrows indicate the change/expansion of area inferred. Biogeographic regions are identified with colors as in Fig. 2: (A) red, W Palearctic; (B) green, North America; (C) cream, South America; (D) clear blue, C-NE Palearctic; (E) yellow, SE Asia; (F) dark blue, Sub-Saharan Africa; (G) pink, Australasia-Oceania. The map of Pleistocene-Holocene period represents the current distribution of each of the nine main subgroups of the subfamily. An animated movie showing the colonization events in sequential order is provided in Appendix S7. 2019, Volume 106 • Otero et al.—Epizoochory coupled with lurking factors boosted diversification • 9

S7). During the middle Miocene, colonization events occurred traits: Lepechiniella inconspicua and Thyrocarpus glochidiatum) in the early Asperugeae, traversing from the Palearctic to North (Fig. 4). America, and the early Myosotideae, spreading from East to West Palearctic. In contrast, contraction of geographic ranges Trait-dependent diversification rates and epizoochory was inferred for Microulinae, from a widespread ancestor in the Palearctic to two geographically differentiated lineages in Africa Results of the 44 averaged models point to higher diversification rates and the C-NE Palearctic. The biogeographic patterns within in lineages with glochids when hidden factors are considered (Fig. 5). Omphalodeae were uncertain since the internal nodes were not The average net diversification for each observed state (with/with- well supported. The remaining between-areas dispersal events oc- out glochids) shows a higher diversification rate mean for lineages curred from the late Miocene to the Pliocene. Remarkably, seven with glochids (Fig. 5B). Nevertheless, there is high rate heterogene- of the nine main clades were able to colonize a considerable num- ity among glochidiate lineages, and part of the diversification rate ber of geographic areas: Rochelieae (five areas), Trichodesmeae overlaps with non-glochidiate lineages (Fig. 5B). The diversification (four areas), Omphalodeae (four areas), Myosotideae (four ar- rates in the mostly glochid-bearing Rochelieae and Cynoglossinae eas), Cynoglossinae (four areas), Asperugeae (three areas), and are significantly higher than in other Cynoglossoideae clades (Fig. 5; Microulinae (three areas) (Fig. 3). Appendices S11, S12), under the joint influence of glochids and hidden factors. Indeed, the influence of other unmeasured factors is also Evolution of diaspore traits revealed at the other four main clades with glochids (South African Microulinae, Adelinia, and Andersonglossum within Amsinckiinae; Our ancestral character-state reconstructions indicated that epi- Suchtelenia calycina and Heterocaryum subsesile within Rochelideae; zoochory was the most likely ancestral dispersal syndrome in Trichodesma africanum, T. calcaratum, and early-diverging lineages Cynoglossoideae, particularly, via glochids as the ancestral an- within Cynoglossinae; Fig. 5) that showed a lower diversification choring structure (Fig. 4; Appendices S8, S9). This epizoochorous rate than expected. The confident assessment of unmeasured fac- trait is unique to Cynoglossoideae among Boraginaceae, and ap- tors (life history, dispersal vector, enviromental conditions) that pears consistently as the ancestral state for the subfamily in five may have been involved is nearly impossible to know; we can only of nine subclades (Trichodesmeae, Rochelieae, Amsinckiinae, hypothesize based on the spatiotemporal frame of the evolution of Microulinae, and Cynoglossinae). Interestingly, glochids appear to Cynoglossoideae (see below). Reconstruction of diversification rates have been maintained for 16 myr (mid-Eocene to mid-Oligocene), through BAMM also identified shifts to higher diversification rates followed by seven independent losses. Additional epizoochorous for both Rochelieae and Cynoglossinae at around the same time traits (uncinated trichomes on nutlets and calyx) were acquired in (20–17 mya; Appendices S11, S12). lineages where glochids persisted and also in clades that lost them (Appendix S9). Anemochorous and myrmecochorous traits have been acquired 14 and four independent times, respectively, since DISCUSSION the late Miocene (Fig. 4). A narrow nutlet margin (an unspecialized trait) was revealed as the ancestral state that later changed to Our results reveal the long existence of the subfamily (1) a wide wing that facilitated the acquisition of anemochorous Cynoglossoideae, dating back to the early Eocene in the C-NE traits, (2) an involute margin related to certain myrmecochorous Palearctic, from which it colonized the W Palearctic, SE Asia, traits, and (3) a loss of margin that can lead to either an epizoo- America, Africa, and Australasia. Each of the nine main clades chorous nutlet if anchoring appendages are maintained or an un- within the Cynoglossoideae has followed a different range ex- specialized nutlet in the absence of any trait related to dispersal pansion history and diversification pattern. No clear patterns are (Appendix S10). Deeply dentate nutlet margin (an epizoochor- inferred between biogeographic shifts and particular dispersal syn- ous trait) was not observed in any ancestral node (Appendix S9). dromes. In particular, fruit glochids were not the only factor that The loss of the ancestral epizoochorous traits was followed by the facilitated colonization and thus triggered diversification rates in gain of anemochorous traits in certain groups of Trichodesmeae, two clades of Cynoglossoideae (Rochelieae and Cynoglossinae). Omphalodeae, Amsinckiinae, and Cynoglossinae. Meanwhile, Additional unmeasured factors related to the biogeographic history the loss of epizoochorous traits and the gain of myrmecochorous of these two clades may have also played a major role in diversifi- traits were inferred in Asperugeae and Bothriosperminae (Fig. 4). cation rates. In Microulinae, the epizoochorous ancestral characters changed into those of an anemochorous syndrome that further shifted to Differentiation of Cynoglossoideae since early Eocene a myrmecochorous syndrome in some lineages (Fig. 4). Two contrasting patterns of epizoochorous trait evolution are inferred for The topology of the phylogenetic reconstruction was congru- Rochelieae and Myosotideae: the former maintained ancestral ent with that of previous studies (Weigend et al., 2013; Chacón epizoochorous traits, while in the latter an unspecialized diaspore et al., 2016). In particular, the nine main clades found (Fig. 2) was the ancestral state that led to the epizoochorous calyx append- agree with the nine major subgroups proposed in the last tax- ages in extant species (Fig. 4). The presence of epizoochorous traits onomic rearrangement of the subfamily (Chacón et al., 2016). (glochids) together with anemochorous traits (wide and open Similarly, our divergence age estimates were congruent with margin) and myrmecochorous traits (mesocarpic protrusion) is those obtained in previous studies (Chacón et al., 2017; Luebert observed in only eight extant species (epizoochorous and anemo- et al., 2017). Early divergence events within Cynoglossoideae chorous traits: Microparacaryum intermedium, Paracaryum ancy- occurred during a globally cooling climate from the mid- ritanum, Pectocarya penicillata, Suchtelenia calycina, Trichodesma Eocene to the Eocene-Oligocene boundary (45–34 mya; Fig. 2; calycosum, and T. africana; epizoochorous and myrmecochorous Liu et al., 2009). Worldwide temperature reduction extending 10 • American Journal of Botany

A B

FIGURE 4. Ancestral character-state reconstructions of Cynoglossoideae using stochastic mapping (SIMMAP). (A) Main diaspore trait reconstruction. (B) Fruit glochid reconstruction. Color legend and representative nutlet images for each state are provided. Diaspore traits reconstructions are used in video animation of Appendix S7. from the late Eocene to part of the Miocene may have been re- Cyperaceae [Escudero et al., 2012]; tribe Mentheae [Drew and sponsible for the extinction of tropical plant species (Prothero, Sytsma, 2012]; tribe Ipomoeeae [Eserman et al., 2014]; family 1994) and concomitant expansion and diversification of tem- Solanaceae [Särkinen et al., 2013]; subfamily Pooideae, Poaceae perate angiosperm groups like Cynoglossoideae (e.g., Carex, [Pimentel et al., 2017]). The early evolution of Cynoglossoideae 2019, Volume 106 • Otero et al.—Epizoochory coupled with lurking factors boosted diversification • 11

A B 0. 8

e .7 0. 60

Net diversification .5

0.44 net.div 0.66 0. 40

No glochids Glochids 0 State 1 FIGURE 5. Reconstruction of diversification rates associated with the presence of glochids through HiSSE. (A) Averaged plot of the diversification rate from the 44 models tested. Color scale from blue (low) to red (high) at the edge of the tree branches indicate the reconstruction of diversification rates. Color legend of diversification rate states is provided. Glochid presence (state 1) is represented by black internal branches, while white internal branches indicate glochid absence (state 0). (B) Net diversification (y axis) is presented for each of the observed states (x axis: no glochids in blue vs. glochids in red). Blue and red horizontal thick lines of each bean plot indicate the mean diversification rate for each state. The short and thin horizontal lines within each bean plot represent each diversification value. The majority of the diversification values lie in the area where the bean plot is widest. during the Eocene-Oligocene occurred in the C-NE Palearctic great colonization success of extant Cynoglossoideae (Fig. 3). This (Fig. 2). This ancestral area differs from that identified in previ- successful colonization pattern contrasts with that of the other ous studies (Luebert et al., 2017: Eurasia-Africa; Chacón et al., two subfamilies of Boraginaceae with narrower distributions that 2017: W Palearctic); however, the widespread ancestor obtained imply inherent colonization limitations (Cohen, 2013). in Luebert et al. (2017) is probably caused by area lumping and LDD has been suggested to account for major intercontinen- limited sampling of Cynoglossoideae (nine of ~50 genera). The tal disjunctions in angiosperms occurring from the Miocene and most restricted W Palearctic was found to be the ancestral area the Holocene (Wen and Ickert-Bond, 2009). Indeed, major inter- for the subfamily, while the delimitation of Eurasia in Chacón continental disjunctions in Cynoglossoideae are inferred since the et al. (2017) lumped together the temperate and cold regions Neogene (Fig. 3). In particular, our biogeographic analysis indi- of north and central-eastern Asia (e.g., Tibet and Mongolia) cates that most Old–New World disjunctions originated between with tropical and subtropical areas of SE Asia (e.g., Vietnam the Miocene and the Pliocene (e.g., Omphalodeae, Asperugeae, and southeast China). Nevertheless, Central Asia (the Irano- Myosotideae, Rochelieae; Fig. 2), which was also observed in the Turanian region and the Tibetan Plateau) is not only regarded as other two subfamilies of Boraginaceae (e.g., Lithospermum and the center of diversity of Cynoglossoideae (Weigend et al., 2013) Anthyphytum; Chacón et al., 2017). but is also considered a major center of diversification of temperate angiosperms and a source of Mediterranean and Eurasian Relationship between diversification rates of Cynoglossoideae temperate flora (Djamali et al., 2012; Manafzadeh et al., 2013; and glochids Wen et al., 2014). The early-Eocene C-NE Palearctic is the accepted center of diversification for two hyperdiverse groups of Structurally complex glochidiate fruits (Selvi et al., 2011; Figs. 1F asterids: the Campanulaceae (~1200 spp.) between 19.9 and 29.7 and 4) appear to be the ancestral condition in the subfamily and mya (Roquet et al., 2009) and the tribe Cardueae (~2400 spp.) have been retained by all subclades. The appearance of glochids in a of Asteraceae between 35.80 and 44.26 mya (Barres et al., 2013). basal-most position of Cynoglossoideae occurred during the early During the Eocene-Oligocene epochs, in situ speciation Eocene (Figs. 2 and 4) when many epizoochorous traits evolved in was revealed as the main process leading to the early diver- angiosperms and species radiation of birds and mammals began gence of five clades of Cynoglossoideae within C-NE Palearctic (Tiffney, 1986). Interestingly, several independent transitions to the (Trichodesmeae, Asperugeae, Myosotideae, Rochelieae, and less complex glabrous and smooth fruits convergently occurred in Cynoglossinae), with the only LDD event detected from C-NE generic lineages within the subfamily in more recent times (e.g., Palearctic to America to account for the origin of Amsinckiinae. Mertensia and Myosotis, among others). From the early Miocene onward, the Cynoglossoideae expanded Many studies have focused on the relationship between fleshy until reaching their current worldwide distribution (Fig. 3). The fruits and diversification rates in angiosperm groups (Biffin biogeographic patterns inferred from our analyses are congruent et al., 2010; Beaulieu and Donoghue, 2013), while few studies with those described by Chacón et al. (2017) and illustrate the have analyzed the effect of dry fruits in shifts of diversification 12 • American Journal of Botany

rates. Subfamily Cynoglossoideae have a high net diversification species of Cynoglossinae that bear barbs, hooks, and glochids rate (0.31; Appendices S11, S12) compared to the mean rate esti- (Selvi et al., 2011). They may have contributed to the spread of mated for the whole family (0.1022; Magallón and Castillo, 2009; Cynoglossum s.l. across the Mediterranean floristic region (Shmida, Table 1). A constant diversification rate in Cynoglossoideae has 1978; Shmida and Ellner, 1983; Montserrat and Alejandre, 2005; been interrupted at least twice due to a significant diversifica- Couvreur et al., 2008). Moreover, the progressive cooling and aridity tion rate increase inferred for Rochelieae and Cynoglossinae of central Asia from the middle Miocene may have contributed to (Fig. 5; Appendices S11, S12). Glochid presence is correlated the expansion and diversification of open grassland habitats in par- with higher diversification rates during the early divergence of allel to the spread of main clades of ruminants (Jacobs et al., 1999; these two distantly related groups (Fig. 5; Appendices S11, S12). Strömberg, 2011). Thus, we hypothesize that the great diversification However, both Rochelieae and Cynoglossinae are characterized of Rochelieae and Cynoglossinae was favored by the diversification by the retention of glochids as a plesiomorphic character state and expansion of the Pecoran ungulates as dispersal vectors, fol- and, thus, no key fruit-dispersal apomorphic innovation seems lowing the spread of large, grass-based, open habitats. Considering to be associated with the diversification rate increase. There are our data in the context of these paleoecological studies, we propose other lineages that retained glochids over time, but their diver- that the diversification and range expansion of their main dispersal sification rates are lower (South African Microulinae,Adelinia , vector may be the unmeasured factors that contributed to the rapid and Andersonglossum within Amsinckiinae; Suchtelenia and Oligocene speciation in Rochelieae and Cynoglossinae. Heterocaryum within Rochelideae; Trichodesmeae and early- diverging lineages within Cynoglossinae; Fig. 5; Appendices S11, S12). Those lower rates in other glochidate lineages indicate that CONCLUSIONS glochid presence was not the only factor that triggered diversification, and that the selective pressures that may have led to This spatiotemporal study reveals a long existence of subfam- rapid speciation through epizoochory were additionally acting ily Cynoglossoideae (early Eocene) in the C-NE Palearctic, from on other, still unidentified factors (Fig. 5; Appendices S11, S12). which it colonized both hemispheres. The nine main clades within These results are congruent with other studies that also observed Cynoglossoideae have different diversification patterns and range shifts in diversification rates in large taxonomic groups resulting expansion histories, although two (Rochelieae and Cynoglossinae) from a suite of factors, not just by the acquisition a key inno- share a similar geographic and temporal window of biogeographic vation (Annonaceae: Erkens et al., 2012; Campanulids: Beaulieu patterns (biogeographic congruence). None of these patterns are and Donoghue, 2013; Bromeliaceae: Silvestro et al., 2014). unequivocally associated with particular LDD syndromes. In situ In agreement with previous studies such as Moore and divergence of only two of six glochid bearing clades (Rochelieae Donoghue (2007), the impact of LDD on diversification rates and Cynoglossinae) in the C-NE Palearctic resulted in a signifi- is suggested to be less important than historically considered. cant increase in diversification rates, indicating that glochid pres- Contrary to our expectations, the successful range expansions of ence alone is not responsible for range expansion and speciation. Cynoglossoideae were not related to increased diversification in Additional factors, such as dispersal vectors, habitat availability, and the majority of the clades. Diversification rates did not increase climate conditions must be invoked to fully account for the histori- in three of the five subgroups of Cynoglossoideae that were the cal diversification events in these groups. most successful colonizers (Trichodesmeae, Omphalodeae, Myosotideae) and most likely to encounter new habitats to drive speciation (Willis et al., 2014; Spriggs et al., 2015). In contrast, ACKNOWLEDGEMENTS the Rochelieae and Cynoglossinae, which have the highest diversification rates, experienced no important geographic changes in The authors thank those botanists that provided plant material, par- early evolutionary histories. ticularly S. Ovchinnikova (Central Siberian Botanical Garden), P. B. A shared geographic (C-NE Palearctic) and temporal (28.60– Heenan (Allan Herbarium), and J. C. Zamora (Uppsala University), 21.59 mya, late Oligocene) window of opportunity for diversifi- as well as the lab and SEM technicians E. Cano and Y. Ruiz from Real cation was found in the origin of Rocheliaeae and Cynoglossinae, Jardín Botánico (Madrid, Spain) and A. J. García from the Museo which is one more example of biogeographic congruence in an- de Ciencias Naturales (Madrid, Spain). Likewise we gratefully ac- giosperms (see review in Vargas et al., 2014). Increases in diversi- knowledge the hard work of the Microscopy Service of Centro de fication rate were observed in the two clades of Cynoglossoideae, Investigación Tecnología e Innovación (CITIUS) at the University in contrast to those inferred for the remaining clades with glochids of Seville (Spain), including a special mention of Prof. J. Arroyo that diversified at a lower rate (Fig. 5: Appendices S11, S12). The cor- (Department of Plant Biology and Ecology at the University of relation between glochid presence and diversification rate increase Seville). We also appreciate the loans provided by numerous herbaria in Rochelieae and Cynoglossinae, when considering unmeasured mentioned in Appendix S1. We are especially grateful to the support factors, is congruent with the contingency conception proposed for and assistance provided by J. Beaulieu, B. O’Meara, I. Sanmartín, M. the Dipsacales: the right diaspore trait must be in the right place at Fernández-Mazuecos, J. V. Sandoval, I. Villa-Machío, Y. Arjona, J. the right time (Moore and Donoghue, 2007). Interestingly, the ori- Fernández-López, and the researchers of the Department of Botany, gin and spread of the main groups of Pecoran ungulates (infraorder University of Pablo de Olavide (Seville). We also thank G. Johnson Pecora, the main group of ruminants) have been inferred around (Smithsonian Institution) for critical help with the language, as well the same time of 27.6–22.4 mya and the same geographic region as the helpful comments of two anonymous reviewers. This study in eastern Eurasia (Hassanin, 2015) as those of the epizoochorous was supported by the Fundación General CSIC and the Banco Rochelieae and Cynoglossinae. Cattle (family Bovidae) have been Santander (universidades) as part of the project titled “Do All identified as one of the main medium-range dispersal vectors of Endangered Species Hold the Same Value? Origin and Conservation 2019, Volume 106 • Otero et al.—Epizoochory coupled with lurking factors boosted diversification • 13

of Living Fossils of Flowering Plants Endemic to Spain” and grant APPENDIX S8. Ancestral state reconstruction (SIMMAP) of three FPU13/04394 from the Spain Ministry of Education. main diaspore traits of Cynoglossoideae (epizoochorous, anemochorous, and myrmecochorous traits). Color legend is provided. AUTHOR CONTRIBUTIONS APPENDIX S9. Ancestral state reconstruction (SIMMAP) of anchoring structures different from gloquids. Color legend is pro- A.O. did the material sampling, analyses, and manuscript prepara- vided for each state. tion as the main author. P. J.-M. participated in material sampling, APPENDIX S10. Ancestral state reconstruction (SIMMAP) of analyses, and manuscript revision. V.V. participated in analyses and margin shape in Cynoglossoideae. manuscript revision. P.V. led the manuscript preparation, writing, and revision. All authors contributed to the experimental design of APPENDIX S11. The nine most probable distinct shift configu- this research. rations sampled with BAMM. The nine configurations account for 94% of the whole probability. Numbers above each plot indicate frequency of shift configurations with the highest posterior probability. DATA ACCESSIBILITY Warmer colors denote faster rates of speciation. Gray circles indicate the locations of rate shifts for each distinct shift configuration. All sequences generated in this study are deposited in GenBank Circle size is proportional to the marginal probability of a shift in (Appendix S1). The nuclotide alignment and tree files used in this each branch. study are available from the TreeBase repository: http://purl.org/ phylo/treebase/phylows/study/TB2:S23831. APPENDIX S12. Speciation rate through time (RTT) plots for subfamily Cynoglossoideae and its main subgroups. Names of the main clades are provided. Labels of x and y axes indicate time be- SUPPORTING INFORMATION fore present (myr) and speciation rate, respectively. Dark line marks speciation rate mean, and shaded range indicates the variance of the Additional Supporting Information may be found online in the speciation rate. Labels of x and y axes are provided. supporting information tab for this article. APPENDIX S13. Maximum clade credibility (MCC) tree from APPENDIX S1. List of voucher specimens and GenBank num- BEAST. The 95% highest posterior density (HPD) bars of diver- bers of species used for phylogenetic study. Asterisks mark acces- gence time are provided for those nodes with ≥0.84 Bayesian pos- sions newly sequenced. Herbarium codes from Index Herbariorum terior probability (BPP). Numbers next to the nodes are the BPP (http://sweetgum.nybg.org/science/ih/) for each voucher are indi- values. Scale axis from 70 myr until present (0 myr) is provided. cated in brackets. APPENDIX S2. Data sources. (A) Taxonomic circumscription, LITERATURE CITED geographic distribution, type of fruit, and syndrome of each of the ingroup species analyzed in this study. For each genus, the propor- Avise, J. C., and B. Riddle. 2009. Phylogeography: Retrospect and prospect. tion of gloquidiate species included over the total number of spe- Journal of Biogeography 36: 3–15. cies with glochids is specified as obtained from the bibliographic Barres, L., I. Sanmartín, C. L. Anderson, A. Susanna, S. Buerki, M. Galbany- sources provided. The species included in the diaspore trait study. Casals, and R. Vilatersana. 2013. Reconstructing the evolution and biogeo- (B) Geographic distribution and main data considered for the out- graphic history of tribe Cardueae (Compositae). American Journal of Botany group. (C) Sources of information used for identification of species 100: 867–882. and fruit. Beaulieu, J. M. 2016. Running HiSSE. https://github.com/thej022214/hisse/tree/ master/vignettes. APPENDIX S3. List of the studied material used for the diaspore Beaulieu, J. M. 2017. Package ‘hisse’. https://cran.r-project.org/web/packages/ trait study. The technique implemented is specified: scanning elec- hisse/index.html. tron microscopy (SEM) and/or microtome stained cross section Beaulieu, J. M., and M. J. Donoghue. 2013. Fruit evolution and diversification in (LM). campanulid angiosperms. Evolution 67: 3132–3144. Beaulieu, J. M., and B. C. O’Meara. 2016. Detecting hidden diversification shifts APPENDIX S4. Summary of logLikelihood (logL), AIC values, Δ in models of trait-dependent speciation and extinction. Systematic Biology AIC, and AIC weights of 44 HiSSE models (q = transition rate, τ = 65: 583–601. turnover rate, ε = extinction fraction). Observed states: 0 = absence Biffin, E., E. J. Lucas, L. A. Craven, I. R. Da Costa, M. G. Harrington, and M. of glochids; 1 = presence of glochids. Hidden states: A = absence of D. Crisp. 2010. Evolution of exceptional species richness among lineages of hidden effect; B, C, D = presence of hidden effects. fleshy-fruited Myrtaceae. Annals of Botany 106: 79–93. Blanco-Pastor, J. L., P. Vargas, and B. E. Pfeil. 2012. Coalescent simulations re- APPENDIX S5. Analysis performed with BAMMtools. veal hybridization and incomplete lineage sorting in Mediterranean Linaria. PLoS One 7: e39089. APPENDIX S6. Phylogenetic reconstruccion of subfamily Burger, W. C. 1981. Why are there so many kinds of flowering plants? BioScience Cynoglossoideae. Numbers on branches indicate phylogenetic 31: 572–581. supports for BEAST, MrBayes, and RaxML analyses, respectively Burkart, M. 2000. River corridor plants (Stromtalpflanzen) in Central European (BEAST/Mr.Bayes/RaxML). lowland: A review of a poorly understood plant distribution pattern. Geologia Sudetica 33: 449–468. APPENDIX S7. Animated movie showing the different coloniza- Chacón, J., F. Luebert, H. H. Hilger, S. Ovchinnikova, F. Selvi, L. Cecchi, C. M. tion events among the areas in sequential order (according the in- Guilliams, et al. 2016. The borage family (Boraginaceae s. str.): A revised inf- ferred mean ages). rafamilial classification based on new phylogenetic evidence, with emphasis 14 • American Journal of Botany

on the placement of some enigmatic genera on the placement of some enig- Hasenstab-Lehman, K. E., and M. G. Simpson. 2012. Cat’s eyes and popcorn matic genera. Taxon 65: 523–546. flowers: Phylogenetic systematics of the genus Cryptantha sl (Boraginaceae). Chacón, J., F. Luebert, and M. Weigend. 2017. Biogeographic events are not cor- Systematic Botany 37: 738–757. related with diaspore dispersal modes in Boraginaceae. Frontiers in Ecology Hassanin, A. 2015. Systematics and Phylogeny of Cattle. In D. J. Garrick and A. and Evolution https://doi.org/10.3389/fevo.2017.00026 DOI. Ruvinsky [eds.], The genetics of cattle, 1–18. CABI, Wallingford, UK. Chandler, M. E. J. 1964. The Lower Tertiary floras of southern England. IV. A Heleno, R., and P. Vargas. 2015. How do islands become green? Global Ecology summary and survey of findings in the light of recent botanical observa- and Biogeography 24: 518–526. tions. British Museum (Natural History), London. Hilger, H. H. 2014. Ontogeny, morphology, and systematic significance Cohen, J. I. 2013. A phylogenetic analysis of morphological and molecular char- of glochidiate and winged fruits of Cynoglosseae and Eritrichieae acters of Boraginaceae: Evolutionary relationships, taxonomy, and patterns (Boraginaceae). Plant Diversity and Evolution 131: 167–214. of character evolution. Cladistics 30: 1–31. Holstein, N., J. Chacon, A. Otero, P. Jiménez-Mejías, and M. Weigend. 2016. Corriveau, J. L., and A. W. Coleman. 1988. Rapid screening method to detect Towards a monophyletic Omphalodes—or an expansion of North American potential biparental inheritance of plastid DNA and results for over 200 an- Mimophytum. Phytotaxa 288: 131–144. giosperms. American Journal of Botany 75: 1443–1458. Jacobs, B. F., J. D. Kingston, and L. L. Jacobs. 1999. The origin of grass-dominated Costa, J. M., J. A. Ramos, L. P. da Silva, S. Timoteo, P. M. Araújo, M. S. Felgueiras, ecosystems. Annals of the Missouri Botanical Garden 86: 590–643. A. Rosa, et al. 2014. Endozoochory largely outweighs epizoochory in migrat- Kadota, Y. 2009. Omphalodes akiensis (Boraginaceae), a new species from ing passerines. Journal of Avian Biology 45: 59–64. Hiroshima prefecture, western Japan. Journal of Japanese Botany 84: Couvreur, M., K. Verheyen, M. Vellend, I. Lamoot, E. Cosyns, M. Hoffmann, 342–349. and M. Hermy. 2008. Epizoochory by large herbivores: Merging data with Kampstra, P., and M. P. Kampstra. 2014. Package ‘beanplot’. models. Basic and Applied Ecology 9: 204–212. Katoh, K., and K. Standley. 2013. MAFFT multiple sequence alignment software Djamali, M., S. Brewer, S. W. Breckle, and S. T. Jackson. 2012. Climatic deter- version 7: Improvements in performance and usability. Molecular Biology minism in phytogeographic regionalization: A test from the Irano-Turanian and Evolution 30: 772–780. region, SW and Central Asia. Flora—Morphology, Distribution, Functional Kearse, M., R. Moir, A. Wilson, S. Stones-Havas, M. Cheung, S. Sturrock, S. Ecology of Plants 207: 237–249. Buxton, et al. 2012. Geneious Basic: An integrated and extendable desk- Donoghue, M. J., and M. J. Sanderson. 2015. Confluence, synnovation, and de- top software platform for the organization and analysis of sequence data. pauperons in plant diversification. New Phytologist 207: 260–274. Bioinformatics 28: 1647–1649. Donoghue, M. J., and S. A. Smith. 2004. Patterns in the assembly of temperate Laenen, B., A. Machac, S. R. Gradstein, B. Shaw, J. Patiño, A. Désamoré, B. forests around the Northern Hemisphere. Philosophical Transactions of the Goffinet, et al. 2016. Increased diversification rates follow shifts to bisexual- Royal Society of London 359: 1633–1644. ity in liverworts. New Phytologist 210: 1121–1129. Drew, B. T., and K. J. Sytsma. 2012. Phylogenetics, biogeography, and staminal Larson-Johnson, K. 2016. Phylogenetic investigation of the complex evolution- evolution in the tribe Mentheae (Lamiaceae). American Journal of Botany ary history of dispersal mode and diversification rates across living and fossil 99: 933–953. Fagales. New Phytologist 209: 418–435. Drummond, A. J., M. A. Suchard, D. Xie, and A. Rambaut. 2012. Bayesian phy- Liu, Z., M. Pagani, D. Zinniker, R. DeConto, M. Huber, H. Brinkhuis, S. R. Shah, logenetics with BEAUti and the BEAST 1.7. Molecular Biology and Evolution et al. 2009. Global cooling during the Eocene-Oligocene climate transition. 29: 1969–1973. Science 323: 1187–1190. Erkens, R. H., L. W. Chatrou, and T. L. Couvreur. 2012. Radiations and key Luebert, F., G. Brokamp, J. Wen, M. Weigend, and H. H. Hilger. 2011. innovations in an early branching angiosperm lineage (Annonaceae; Phylogenetic relationships and morphological diversity in Neotropical Magnoliales). Botanical Journal of the Linnean Society 169: 117–134. Heliotropium (Heliotropiaceae). Taxon 60: 663–680. Escudero, M., A. L. Hipp, M. J. Waterway, and L. M. Valente. 2012. Diversification Luebert, F., T. L. P. Couvreur, M. Gottschling, H. H. Hilger, J. S. Miller, and M. rates and chromosome evolution in the most diverse angiosperm genus Weigend. 2017. Historical biogeography of Boraginales: West Gondwanan of the temperate zone (Carex, Cyperaceae). Molecular Phylogenetics and vicariance followed by long-distance dispersal? Journal of Biogeography 44: Evolution 63: 650–655. 158–169. Eserman, L. A., G. P. Tiley, R. L. Jarret, J. H. Leebens-Mack, and R. E. Miller. Magallón, S., and A. Castillo. 2009. Angiosperm diversification through time. 2014. Phylogenetics and diversification of morning glories (tribe Ipomoeeae, American Journal of Botany 96: 349–365. Convolvulaceae) based on whole plastome sequences. American Journal of Magallón, S., S. Gómez-Acevedo, L. L. Sánchez-Reyes, and T. Hernández- Botany 101: 92–103. Hernández. 2015. A metacalibrated time-tree documents the early rise of Gabel, M. 1987. A Fossil Lithospermum (Boraginaceae) from the Tertiary of flowering plant phylogenetic diversity. New Phytologist 207: 437–453. South Dakota. America 74: 1690–1693. Manafzadeh, S., G. Salvo, and E. Conti. 2013. A tale of migrations from east to Gabel, M. L., D. C. Backlund, and J. Haffner. 1998. The Miocene macroflora west: The Irano-Turanian floristic region as a source of Mediterranean xero- of the northern Ogallala Group, northern Nebraska and southern South phytes. Journal of Biogeography 41: 366–379. Dakota. Journal of Paleontology 72: 388–397. Melcher, I. M., F. Bouman, and A. M. Cleef. 2000. Seed dispersal in paramo Gorb, E. V., and S. N. Gorb. 2003. Seed dispersal by ants in a deciduous for- plants: Epizoochorous and hydrochorous taxa. Plant Biology 2: 40–52. est ecosystem: Mechanisms, strategies, adaptations. Kluwer Academic, Meyer, A. L., and J. J. Wiens. 2018. Estimating diversification rates for higher Dordrecht, The Netherlands. taxa: BAMM can give problematic estimates of rates and rate shifts.Evolution Gorb, S. N., E. V. Gorb, and P. Punttila. 2000. Effects of re-dispersal of seeds 72: 39–53. by ants on the vegetation pattern in a deciduous forest: A case study. Acta Miller, M. A., W. Pfeiffer, and T. Schwartz. 2010. Creating the CIPRES Science Oecologica 21: 293–301. Gateway for inference of large phylogenetic trees. Proceedings of the Gateway Graham, A., and D. M. Jarzen. 1969. Studies in Neotropical paleobotany. I. The Computing Environments Workshop 1–8. Oligocene communities of Puerto Rico. Annals of the Missouri Botanical Montserrat, P., and J. A. Alejandre. 2005. Los Cynoglossum “germanicum, pustu- Garden 56: 308–357. latum and dioscoridis” pirenaico-cantabricos. Bulletin de la Société d’histoire Gürke, M. 1893. Borraginaceae (Asperifoliaceae). In A. Engler and K. Prantl naturelle de Toulouse 141: 31–35. [eds.], Die Natürlichen Pflanzenfamilien, 71–131. W. Engelmann, Leipzig. Moore, B. R., and M. J. Donoghue. 2007. Correlates of diversification in the plant Hammouda, S. A., M. Weigend, F. Mebrouk, J. Chacón, M. Bensalah, H. J. clade Dipsacales: Geographic movement and evolutionary innovations. The Ensikat, and M. Adaci. 2015. Fossil nutlets of Boraginaceae from the conti- American Naturalist 170: S28–S55. nental Eocene of Hamada of Méridja (southwestern Algeria): The first fossil Moore, B. R., S. Höhna, M. R. May, B. Rannala, and J. P. Huelsenbeck. 2016. of the Borage family in Africa. American Journal of Botany 102: 2108–2115. Critically evaluating the theory and performance of Bayesian analysis 2019, Volume 106 • Otero et al.—Epizoochory coupled with lurking factors boosted diversification • 15

of macroevolutionary mixtures. Proceedings of the National Academy of Segal, R. 1966. Taxonomic study of the fossil species of the genus Cryptantha Sciences USA 113: 9569–9574. (Boraginaceae). The Southwestern Naturalist 11: 205–210. Otero, A., P. Jiménez-Mejías, V. Valcárcel, and P. Vargas. 2014. Molecular phylo- Selvi, F., A. Coppi, and L. Cecchi. 2011. High epizoochorous specializa- genetics and morphology support two new genera (Memoremea and Nihon) tion and low DNA sequence divergence in mediterranean Cynoglossum of Boraginaceae s.s. Phytotaxa 173: 241–277. (Boraginaceae): Evidence from fruit traits and ITS region. Taxon 60: Oxelman, B., M. Lidén, and D. Berglund. 1997. Chloroplast rps16 intron phylog- 969–985. eny of the tribe Sileneae (Caryophyllaceae). Plant Systematics and Evolution Shmida, A. 1978. Relationships between Paracaryum intermedium and P. bois- 206: 393–410. sieri on Sinai, and the generic separation of Paracaryum and Mattiastrum Paradis, E., J. Claude, and K. Strimmer. 2004. APE: Analyses of phylogenetics (Boraginaceae). Plant Systematics and Evolution 129: 323–326. and evolution in R language. Bioinformatics 20: 289–290. Shmida, A., and S. Ellner. 1983. Seed dispersal on pastoral grazers in open Pimentel, M., M. Escudero, E. Sahuquillo, M. Á. Minaya, and P. Catalán. 2017. Mediterranean chaparral, Israel. Israel Journal of Botany 32: 147–159. Are diversification rates and chromosome evolution in the temperate grasses Silvestro, D., G. Zizka, and K. Schulte. 2014. Disentangling the effects of key in- (Pooideae) associated with major environmental changes in the Oligocene- novations on the diversification of Bromelioideae (Bromeliaceae). Evolution Miocene? PeerJ 5: e3815. 68: 163–175. Plants, J. G. 2014. JSTOR Global Plants. http://plants.jstor.org/. Simpson, M. G. 2012. Cryptantha s.l. taxonomy & images. http://www.sci.sdsu. Popov, M. G. 1953. Boraginaceae. In S. B. K. [ed.], Flora SSSR, vol. 19, 97–718. edu/plants/cryptantha/. Izdatel’stvo Akademii Nauk SSSR, Moskva-Leningrad. Sorensen, A. E. 1986. Seed dispersal by adhesion. Annual Review of Ecology and Posada, D. 2008. jModelTest: Phylogenetic model averaging. Molecular Biology Systematics 17: 443–463. and Evolution 25: 1253–1256. Spriggs, E. L., W. L. Clement, P. W. Sweeney, S. Madriñán, E. J. Edwards, and M. Prothero, D. R. 1994. The late Eocene-Oligocene extinctions. Annual Review of J. Donoghue. 2015. Temperate radiations and dying embers of a tropical past: Earth and Planetary Sciences 22: 145–165. The diversification of Viburnum. New Phytologist 207: 340–354. Rabosky, D. L. 2014. Automatic detection of key innovations, rate shifts, and Stamatakis, A. 2014. RAxML version 8: A tool for phylogenetic analysis and diversity-dependence on phylogenetic trees. PLoS One 9: 1–15. post-analysis of large phylogenies. Bioinformatics 30: 1312–1313. Rabosky, D. L. 2018. BAMM at the court of false equivalency: A response to Strömberg, C. A. E. 2011. Evolution of grasses and grassland ecosystems. Annual Meyer and Wiens. Evolution 72: 2246–2256. Review of Earth and Planetary Sciences 39: 517–544. Rabosky, D. L., J. S. Mitchell, and J. Chang. 2017. Is BAMM flawed? Theoretical Thorne, R. F. 1972. Major disjunctions in the geographic ranges of seed plants. and practical concerns in the analysis of multi-rate diversification models. The Quaterly Review of Biology 47: 365–411. Systematic Biology 66: 477–498. Tiffney, B. H. 1986. Evolution of seed dispersal syndromes according to the fos- Rambaut, A., M. Suchard, D. Xie, and A. Drummond. 2014. Tracer version 1.6. sil record. In D. R. Murray [ed.], Seed dispersal, 273–305. Academic Press, http://beast.bio.ed.ac.uk/Tracer. Sidney, Australia. Raven, P. H. 1972. Plant species disjunctions: A summary. Annals of the Missouri Tiffney, B. H. 2004. Vertebrate dispersal of seed plants through time. Annual Botanical Garden 59: 234–246. Review of Ecology, Evolution, and Systematics 35: 1–29. R Core Team. 2012. R: A language and environment for statistical computing, Vargas, P., B. G. Baldwin, and L. Constance. 1999. A phylogenetic study of version 3.2.3. https://www.r-project.org/. Sanicula sect. Sanicoria and S. sect. Sandwicenses (Apiaceae) based on nu- Ree, R., and I. Sanmartín. 2018. Conceptual and statistical problems with the clear rDNA and morphological data. Systematic Botany 24: 228–248. DEC+ J model of founder-event speciation and its comparison with DEC via Vargas, P., L. M. Valente, J. L. Blanco-Pastor, I. Liberal, B. Guzmán, E. Cano, model selection. Journal of Biogeography 45: 741–749. A. Forrest, and M. Fernández-Mazuecos. 2014. Testing the biogeo- Ree, R., and S. Smith. 2008. Lagrange: software for likelihood analysis of geo- graphical congruence of palaeofloras using molecular phylogenetics: graphic range evolution. Systematic Biology 57: 4–14. Snapdragons and the Madrean-Tethyan flora. Journal of Biogeography 41: Regal, P. J. 1977. Ecology and evolution of flowering plant dominance. Science 932–943. 196: 622–629. Weigend, M., F. Luebert, F. Selvi, G. Brokamp, and H. H. Hilger. 2013. Multiple Renner, S. S. 2005. Relaxed molecular clocks for dating historical plant dispersal origins for hound’s tongues (Cynoglossum L.) and navel seeds (Omphalodes events. Trends in Plant Science 10: 550–558. Mill.)—the phylogeny of the borage family (Boraginaceae s.str.). Molecular Revell, L. J. 2012. Phytools: an R package for phylogenetic comparative biology Phylogenetics and Evolution 68: 604–618. (and other things). Methods in Ecology and Evolution 3: 217–223. Wen, J., and S. M. Ickert-Bond. 2009. Evolution of the Madrean-Tethyan dis- Ronquist, F., M. Teslenko, P. Van Der Mark, D. L. Ayres, A. Darling, S. Höhna, junctions and the North and South American amphitropical disjunctions in B. Larget, et al. 2012. MrBayes 3.2: Efficient Bayesian phylogenetic infer- plants. Journal of Systematics and Evolution 47: 331–348. ence and model choice across a large model space. Systematic Biology 61: Wen, J., J. Zhang, Z. L. Nie, Y. Zhong, and H. Sun. 2014. Evolutionary diver- 539–542. sifications of plants on the Qinghai-Tibetan Plateau. Frontiers in Genetics Roquet, C., I. Sanmartín, N. Garcia-Jacas, L. Sáez, A. Susanna, N. Wikström, 5: 4. and J. J. Aldasoro. 2009. Reconstructing the history of Campanulaceae with Willis, C. G., J. C. Hall, R. Rubio De Casas, T. Y. Wang, and K. Donohue. 2014. a Bayesian approach to molecular dating and dispersal–vicariance analyses. Diversification and the evolution of dispersal ability in the tribe Brassiceae Molecular Phylogenetics and Evolution 52: 575–587. (Brassicaceae). Annals of Botany 114: 1675–1686. Särkinen, T., L. Bohs, R. G. Olmstead, and S. Knapp. 2013. A phylogenetic Zhu, G. L., H. Riedl, and R. V. Kamelin. 1995. Boraginaceae. In Z. Y. Wu and P. framework for evolutionary study of the nightshades (Solanaceae): A dated H. Raven [eds.], Flora of China, vol. 16, 329–427. Science Press, Beijing, and 1000-tip tree. BMC Evolutionary Biology 13: 1–15. Missouri Botanical Garden, St. Louis, Missouri, USA.

View publication stats