ORIGINAL ARTICLE

Diversification of Caiophora (Loasaceae subfam. Loasoideae)during the uplift of the Central Andes

Marina Micaela Strelin1& José Ignacio Arroyo2,3 &

Stella Fliesswasser5 & Markus Ackermann4,5

Received: 25 May 2016 /Accepted: 16 November 2016 /Published online: 29 November 2016# Gesellschaft für Biologische Systematik 2016

Abstract Andean orogeny and the ecological changes thatfollowed promoted diversification in plant and animal line-ages since the Early Miocene. The angiosperm genusCaiophora (Loasaceae, subfam. Loasoideae) comprisesaround 50 species that are endemic to South America. Theseare distributed from southern Ecuador to Central Chile andArgentina. Bee pollination and distribution at low-intermediate elevations probably represent the ancestral con-dition in the lineage that includes Caiophora and its alliedgenera. The majority of Caiophora species grow at high ele-vations in the Andes, where some depend on vertebrate polli-nation. Previous studies did not resolve phylogenetic relation-ships within Caiophora, which precluded the dating of theorigin and divergence of this group. We used markers of onenuclear (ITS) and one plastid region (trnSGCU-trnGUUC) to

solve phylogenetic relationships among 19Caiophora species(including different accessions). We also included 10 speciesof the allied genera Blumenbachia and Loasa. Aosa rostrataand Xylopodia klaprothioides were used as outgroups.Phylogenetic reconstruction strongly supports the monophylyof Caiophora, and although several clades within this genusare poorly supported, our study yielded a better infra-genericresolution than previous studies. The origin of Caiophora isdated to the Early-Middle Miocene and can be related to theuplift of the Cordilleras Frontal and Principal and to Pacificmarine transgressions. According to our estimations,Caiophora began to diversify during the Middle-LateMiocene and this unfolding proceeded eastwards during thePliocene and the Pleistocene, in parallel to the uplift of differ-ent Andean mountain ranges.

Keywords Diversification .Caiophora . Loasaceae, subfam.Loasoideae . Diversification . Central Andes . Orogeny

Introduction

The diversification of a lineage can be promoted by intrinsicfactors, e.g. the acquisition of an evolutionary novelty(Minelli and Fusco 2012), or by extrinsic factors. The lattercan be either abiotic, e.g. orogeny and climate change (Brysonet al. 2012; Solà et al. 2013; Meng and Zhang 2013), or biotic,e.g. changes in the selective regimes (Nylin et al. 2014), andemergence of new ecological niches (Muschick et al. 2012).Among the most striking examples of lineage diversificationare those related to Andean orogeny (Gamble et al. 2008;Antonelli et al. 2009; Chaves et al. 2011; Drummond et al.2012a).

The Andean uplift started about 40 Ma during the MiddleEocene, but the most pronounced uplift events took place in

Electronic supplementary material The online version of this article(doi:10.1007/s13127-016-0312-4) contains supplementary material,which is available to authorized users.

* Marina Micaela Strelinmarina.strelin85@gmail.com

1 Laboratorio Ecotono, INIBIOMA (Universidad Nacional delComahue-CONICET), Pasaje Gutierrez 1125, 8400 Bariloche, RioNegro, Argentina

2 Departamento de Ecología, Facultad de Ciencias Biológicas,Pontificia Universidad Católica de Chile, Santiago, Chile

3 Instituto de Ecología & Biodiversidad (IEB-Chile), Casilla 653,Santiago, Chile

4 Nees Institut für Biodiversität der Pflanzen, RheinischeFriedrich-Wilhelms-Universität, Meckenheimer Allee 170,53115 Bonn, Germany

5 Institut für Integrierte Naturwissenschaften – Biologie, UniversitätKoblenz-Landau, Universitätsstr. 1, 56070 Koblenz, Germany

Org Divers Evol (2017) 17:29–41DOI 10.1007/s13127-016-0312-4

the Late Oligocene-Middle Miocene (∼25 to 17 Ma), in theLate Miocene-Early Pliocene (∼9 to 5 Ma) and in the EarlyPliocene-Pleistocene (∼5–less than 5 Ma) (Lamb et al. 1997).These events are linked to considerable diversification in an-giosperm lineages, e.g. Smith and Baum (2006), Särkinenet al. (2012), Drummond et al. (2012a), Baranzelli et al.(2014) and Moré et al. (2015). In addition to acting as a dis-persion corridor for plant lineages, the rising Andes resulted inthe formation of vicariant barriers fostering allopatric specia-tion in lineages growing on the eastern and the westernAndean slopes or inhabiting different inner Andean valleys(Luebert and Weigend 2014). Glaciations, which resultedfrom the interplay between the Andean uplift and periods ofintense cooling (Mercer and Sutter 1982; Clapperton 1983;Zech et al. 2009; Rubioldo et al. 2001; Martini et al. 2012),constituted additional vicariant barriers between the middleMiocene and the Pliocene and during the Pleistocene (e.g.Cosacov et al. 2010; Sérsic et al. 2011). Glaciations also fos-tered northward dispersal and displacement of different plantlineages (Palazzesi et al. 2012).

The rising of the Andes fostered plant lineage diversi-fication not only through the generation of vicariant bar-riers, but also through the generation of high-elevationareas, which were vacant for colonization, e.g. high-Andean species of Lupinus (Hughes and Eastwood2006). Besides the origin of these high-elevation areas,unprecedented ecological conditions in these new envi-ronments may have also contributed to plant lineage di-versification (Antonelli and Sanmartín 2011; Särkinenet al. 2012; Luebert and Weigend 2014). For instance,while altitude is believed to act as an environmental filteron the diversity of bee communities (Hoiss et al. 2012)and insect pollination tends to be negatively affected byaltitude in the Andes (Arroyo et al. 1985), hummingbirdsdiversified in a context of Andean uplift since the middleMiocene (∼15 Ma) (McGuire et al. 2014). Furthermore,some of these hummingbird lineages, e.g. Ensiferaensifera and Patagona gigas, adapted their flight dynam-ics and their body size to high-elevation conditions in theAndes (Altshuler et al. 2004). The Andean uplift alsofostered diversification in other vertebrate lineages, e.g.small rodents of the Sigmodontinae subfamily (Martinezet al. 2012). In fact, there is evidence that some rodentlineages in this subfamily, e.g. Graomys griseoflavus(Cavides-Vidal et al. 1987) and Phyllotis xanthopygus(Nespolo et al. 1999), evolved adaptations to resist theextreme environmental conditions characteristic of high-Andean environments. Since transitions from insect tohummingbird pollination are likely to take place in envi-ronments where insects are poorly represented (Cruden1972), adaptation to hummingbird pollination, and even-tually to pollination by other vertebrates, may have beenan important factor in driving lineage diversification in

high-Andean environments. The recent origin of a two-species clade in Schizanthus (Solanaceae) from a presum-ably bee-pollinated ancestor is related to plant specializa-tion on hummingbird pollination in high-Andean environ-ments (Perez et al. 2006).

Loasaceae subfamily Loasoideae is a monophyletic lineagelargely restricted to Central and South America. It comprises13 genera and more than 220 species, being the most species-rich group within Loasaceae.NasaWeigend, the largest genuswithin the subfamily, has its centre of diversity in theAmotape-Huancabamba Zone (northern Peru) (Mutke et al.2014). The second largest genus, Caiophora C. Presl(Fig. 1), shows its highest diversity in the area from northernPeru to northern Argentina. The ca. 50 species in this genusare distributed from Ecuador to Central Chile and Argentina,between 2000 and 5000 m a.s.l. This genus is associated todifferent Andean mountain ranges, and only a single species,Caiophora arechavaletae Urb. & Gilg., is found in southernBrazilian and Uruguayan grasslands (Zuloaga et al. 2008).Some Caiophora species, e.g. Caiophora cirsiifolia C. Presl,present phenotypic variants, which were defined on the basisof flower characteristics, e.g. corolla shape and nectar compo-sition, as well as inflorescence and vegetative traits(Ackermann and Weigend 2006).

Bee pollination is presumably the ancestral condition inCaiophora and its allied genera (Strelin et al. 2016a). In addi-tion to bee-pollinated species, Caiophora also includes spe-cies that rely on hummingbird pollination (Ackermann andWeigend 2006; Weigend et al. 2004) as well as a singlerodent-pollinated species, Caiophora coronata (pollinatedby the small rodent G. griseoflavus) (Cocucci and Sérsic1998). Despite the absence of pollinator records forCaiophora pentlandii (Paxton ex Graham) G. Don exLoudon, this species is also believed to be pollinated by smallrodents, based on its floral and inflorescence morphology(Ackermann and Weigend 2006). Previous work showed thatflower shape inCaiophora and allied taxa evolved in responseto bee, hummingbird and rodent pollination (Strelin et al.2016a).

Although the monophyly of Caiophora was demonstratedin previous phylogenetic analyses (Hufford et al. 2003, 2005;Weigend et al. 2004), these studies included a restricted sam-pling of this genus (10 out of the ∼50 Caiophora species inHufford et al. 2005). Phylogenetic relationships between spe-cies of Caiophora are poorly resolved, presumably becauseprevious phylogenetic inferences were based only on partialplastid DNA evidence. This impedes dating the origin ofCaiophora and the diversification events within the genus,which is crucial to better understand the effects that orogenic,climatic and ecological factors may have had on the evolutionof this genus.

In this study, we present a phylogenetic reconstruction ofthe genus Caiophora, including 19 species (with four

30 Strelin M.M. et al.

phenotypic variants of C. cirsiifolia and two of Caiophoracarduifolia). We used both nuclear and plastid DNA evidence.Our specific aims are (1) to explore whether monophyly ofCaiophora holds after estimating a phylogeny that includesDNA samples and more species of the genus, (2) to solvephylogenetic relationships within Caiophora, (3) to date both

the origin of Caiophora and the divergence events within thisgenus and (4) to reconstruct the evolution of pollinationmodes in the studied species. Finally, the results obtained forCaiophora were compared to the results of other studies ad-dressing diversification in South American angiosperm line-ages during Andean uplift.

Fig. 1 Flower diversity in Caiophora (Loasaceae, subfam. Loasoideae)and in allied lineages. a Loasa acerifolia Dombey ex Juss., b L. bergiiHieron., c Caiophora arechavaletae (Urb.) Urb. & Gilg, d C. cernua(Griseb.) Urb. & Gilg ex Kurtz, e C. stenocarpa Urb. & Gilg., fC. cirsiifolia C. Presl (Arequipa form), g C. carduifolia C. Presl (Cuzcoform), h C. carduifolia C. Presl (Apurimac form), i C. cirsiifolia C. Presl

(Ancash/Huaraz form), j C. canarinoides (Lenné & K. Koch) Urb. &Gilg., k C. hibiscifolia (Griseb.) Urb. & Gilg., l C. lateritia Klotzsch, mC. coronata (Gillies ex Arn.) Hook. & Arn., n C. pentlandii (Paxton exGraham) G. Don ex Loudon, o C. chuquitensis (Meyen) Urb. & Gilg andp C. clavata Urb. & Gilg. Scale bar equals 10 mm

Diversification of Caiophora (Loasaceae subfam. Loasoideae) 31

Materials and methods

Taxon sampling and outgroup selection

A total of 19 species of Caiophora (including four forms ofC. cirsiifolia and two forms of C. carduifolia) were includedin this study (Supplementary file 1: Table S1). The samplingof the species was aimed at having a good coverage of thedistribution range of Caiophora (Supplementary file 2:Table S2) and of the pollination modes presented by speciesin this genus, i.e. bee, hummingbird and small rodent pollina-tion (Ackermann and Weigend 2006; 2012; Supplementaryfile 3: Table S3). These pollination modes are in turnassociated with different flower morphologies (Strelin et al.2016a). Studied species belong to 9 out of the 10 infra-genericCaiophora groups, which were defined on the basisof morphological characters by Ackermann (2012), i.e.C. arechavaletae group, Caiophora nivalis group, Caiophorapterosperma group,Caiophora chuquitensis group, Caiophoracontorta group, C. coronata group, Caiophora clavata group,C. cirsiifolia group, Caiophora lateritia group and C. clavatagroup. The only group that we were not able to include was theC. contorta group, composed of two species (Ackermann2012). The C. chuquitensis group is poorly represented, sincewe were able to include only a single representative,C. chuquitensis (Meyen Urb. & Gilg), of this eight-speciesgroup. Furthermore, eight species of Loasa Adans. and threespecies of Blumenbachia Schrad, representing different infra-generic groups, were included. Aosa rostrata (Urb.) Weigendand Xylopodia klaprothioides Weigend were used asoutgroups.

Sequence data

We used the DNeasy Plant Mini Kit (Qiagen, Hilden,Germany) to extract total genomic DNA from silica-dried leafmaterial. The internal transcribed spacer of the nuclear ribo-somal repeat (ITS1 and ITS2) and the trnSGCU-trnGUUC

intergenic spacer were amplified using the following proto-cols: For ITS1 and ITS2, we used the forward primerITS5_bryo (Stech et al. 2013) and the reverse primerITS4_bryo (Stech et al. 2013). Occasionally, we used the in-ternal forward primer SEQITS2_Angio (5 ′-AACGACTCTCGGCAACGG-3′) and the internal reverse primerSEQITS1_Angio (5′-TTGCGTTCAAAGACTCGATGG-3′).Polymerase chain reactions (PCRs) were done according tothe protocol of Wen and Zimmer (1996). For trnSGCU-trnGUUC, we used the forward primer trnS and the reverseprimer trnG (Hamilton 1999). All PCR reactions used 1×GoTaq®Flexi buffer (Promega GmbH, Mannheim,Germany), 0.75 units of GoTaq® Flexi DNA polymerase(Promega GmbH, Mannheim, Germany), 0.05 nmol of for-ward and reverse primers, 0.1 mM of each dNTP, 1 mM of

MgCI2 and 1 μl of 1:20 diluted freshly prepared genomicDNA, and a final volume of 25 μl was obtained with theaddition of ddH2O. The PCR for trnSGCU-trnGUUC was initi-ated with a 95 °C denaturing cycle for 5 min, followed by34 cycles of 95 °C for 30 s, 50 °C for 1 min and 72 °C for90 s, and the reaction was ended with an extension cycle of72 °C for 4 min. PCR amplicons of ITS1 and ITS2 andtrnSGCU-trnGUUC were sequenced in both directions byGATC (Constance, Germany) with the same primers usedfor amplification. Forward and reverse sequences were exam-ined, compared and corrected using the program PhyDe(Müller et al. 2010). The GenBank accession numbers of themarkers included in this study are shown in Supplementaryfile 1: Table S1.

Alignment and phylogenetic reconstruction

Alignments were performed using the program Mafft ver-sion 7 (Katoh and Standley 2013), with the default strat-egy L-INS-i. To test for substitution saturation in themarkers used in the phylogenetic analyses described inthe following, we employed the substitution saturationapproach of Xia et al. (2003). This approach consists incalculating an index of substitution saturation (ISS) for aset of markers, which represents the observed value ofsubstitution saturation. ISS is then statistically comparedto a critical saturation value (ISS.C), beyond which themarkers start failing to recover the correct tree (Xiaet al. 2003). In practice, if ISS is significantly lower thanISS.C, saturation is assumed to be low enough to recon-struct the phylogenetic history of a lineage using that setof markers. Technical details about this approach can beconsulted in Xia et al. (2003). Furthermore, we visually ex-amined the relationship between the uncorrected genetic dis-tances in our alignment and the distances corrected with theF84 model of DNA substitution (Felsenstein 1984).Uncorrected and corrected distances are expected to scale lin-early if there is no substitution saturation. Saturation analyseswere done with the software DAMBE (Xia and Xie 2001).

We estimated phylogenetic relationships among Loasaceaespecies based on a combined ITS/ trnSGCU-trnGUUC align-ment and on each separate marker, using Bayesian inference(BI) as implemented in MrBayes v3.1.2 (Ronquist andHuelsenbeck 2003). The program was used to simultaneouslyestimate tree topology and parameter values for an indepen-dent GTR+Γ model of nucleotide substitution. Two simulta-neous and independent runs were performed for 10 × 106 it-erations of a Markov Chain Monte Carlo (MCMC) algorithm.Each run included six simultaneous chains, sampling every1000 generations. Support for the nodes and parameter esti-mates were derived from a majority-rule consensus of the last5000 trees sampled after convergence. The average standard

32 Strelin M.M. et al.

deviation of split frequencies remained <0.01 after the burn-inthreshold.

To account for the sensitivity of the marker and the recon-struction method, we performed separate phylogenetic recon-structions for the ITS and the trnSGCU-trnGUUC alignments,using alternative methods of phylogenetic reconstruction:maximum parsimony (MP), neighbour joining (NJ) and max-imum likelihood (ML), as implemented in the programsPAUP version 4 (for MP and NJ) (Swofford 2003) andTreefinder version March 2011 (Jobb et al. 2004), respective-ly. For MP, a heuristic search, saving up to 1000 trees, wasperformed, using the branch-swapping BTree bisection andreconnection^ (TBR) algorithm. For NJ, a GTR+Γ modelwas used, with a gamma shape parameter of 0.5 and five ratecategories. For ML, we searched for the best fitting model foreach marker and used a partition when estimating the phylog-eny with the concatenated alignment. The best fitting modelswere TN+Γ (for ITS) and TVM+Γ (for trnSGCU-trnGUUC).

We finally obtained consensus trees for ITS, trnSGCU-trnGUUC and concatenated markers, in order to summarizethe tree topologies recovered with the four methods of phylo-genetic reconstruction. Although we based our conclusions onthe concatenated alignment using BI, since this is a highlyconsistent and efficient method of phylogenetic reconstruction(Yang and Rannala 2012), we also briefly discussed the topol-ogies obtained with separate markers, using other approachesof phylogenetic reconstruction.

Molecular clock test and molecular dating

Before dating, we first tested concatenated markers for theexistence of a molecular clock, using the test of molecularclock implemented in MEGA version 5.2.2 (Tamura et al.2011). Three calibration points were employed for moleculardating, which were obtained from a previous study that in-cluded species of the Loasaceae family (Schenk and Hufford2010): the common ancestor of C. chuquitensis (Meyen) Urb.& Gilg. (Caiophora macrocarpa in that study) and A. rostrata(40.21 Ma [16.95–50.43]), the common ancestor ofC. chuquitensis and Loasa heterophylla Hook & Arn.(28.8 Ma [12.92–40.62]) and the common ancestor ofC. chuquitensis and Loasa pinnatifida Gillies ex Arn.(Loasa filicifolia in that study) (22.53 Ma [9.97–36.11]). Thenumbers in square brackets are the minimum and the maxi-mum of the probability distribution of estimated ages inSchenk and Hufford (2010). Bayesian posterior probabilityfor the first node was 0.59 and above 0.95 for the remainingtwo.

We used three molecular dating methods to estimatethe divergence time among species of our phylogeny.The obtained ML best tree topology, with a 0.0001 minimumedge length and 1000 time replicates, was used both for localrate minimum deformation (LRMD) (Jobb et al. 2004) and

non-parametric rate smoothing (NPRS) (Jobb et al. 2004;Sanderson 1997). Briefly, while NPRS maintains the rates ateach node (ri), as similar as possible to their respective ances-tral rates (r(A)i), LRMD maintains real rates as similar as pos-sible to ideal local rates, accounting for the similarity of ratesamong sister groups. For NPRS and LRMD, we usedTreefinder version March 2011 (Jobb et al. 2004). The sec-ondary calibration points remain fixed in the LRMD and theNPRS estimations. Since the calibration point correspondingto the common ancestor of C. chuquitensis and L. pinnatifidarepresents the origin ofCaiophora, this event was not estimat-ed with LRMD and NPRS. BEAST v1.8.1 (Drummond et al.2012b) was used in the Bayesian approach. Secondary cali-bration points were included in the form of a normal distribu-tion with a standard deviation of 10 Ma, to account for therange of the distribution estimated in Schenk and Hufford(2010). The Bayesian MCMC analysis was run for50.0 × 106 iterations, using a GTR substitution model (empir-ical base frequencies), a gamma site heterogeneity model withfour gamma categories, an exponential log-normal relaxedclock (default) and a Yule speciation process tree prior. Wesampled every 1000 generations, which yielded a total of50,000 trees, of which the first 50% was subsequentlydiscarded. The maximum posterior credibility tree was there-fore constructed with the remaining 25,000 trees. TheBEAUTI file is provided as supplementary information(Supplementary file 4: Appendix S1).

Reconstruction of the evolution of pollination modes

We obtained pollinator data from Ackermann and Weigend(2006), Weigend and Gottschling (2006) and other litera-ture (Supplementary file 3: Table S3). A pollination mode(hummingbird, bee and rodent pollination) was assigned toall species based on these records. Species with ambiguouspollinator records, i.e. reported to be pollinated by hum-mingbirds and bees, C. chuquitensis, C. lateritia andCaiophora hibiscifolia, were assigned either to bee or tohummingbird pollination, based on the study of Strelinet al. (2016a). We pruned the maximum posterior credibil-ity tree to the 25 species for which pollinator data wasavailable. To reconstruct the history of pollination modesin the studied species, we conducted stochastic charactermappings (Nielsen 2002) using the make.simmap functionfrom the R package phytools (Revell 2012). Given a phy-logeny and discrete character states for extant terminaltaxa, this Bayesian method applies a Monte Carlo algo-rithm to sample the posterior probability distribution ofancestral states on the branches of a phylogeny under aMarkov process of evolution. One thousand stochasticcharacter mapping simulations were performed on thepruned maximum posterior credibility BEAST tree, andthe posterior probability of each character state plotted on

Diversification of Caiophora (Loasaceae subfam. Loasoideae) 33

the corresponding node (Revell 2014), using thedescribe.simmap function from the R package phytools(Revell 2012).

Results

Phylogenetic relationships

In the saturation test, the Xia index showed significant nosaturation of markers (Supplementary file 5: Table S4) andsaturation plots showed a linear relationship between the ob-served and the marker distances expected under the F84 mod-el (Supplementary file 6: Fig. S1). This suggests that themarkers in our data set can be used to reconstruct the evolu-tionary history of the species in this study.

Although phylogenetic topologies varied when usingdifferent markers and methods of phylogenetic reconstruc-tion (Supplementary files 7–9: Figs. S2, S3 and S4), allphylogenetic reconstructions recovered the monophyly ofCaiophora (Supplementary file 10: Fig. S5), suggesting asingle origin for this group. Despite the fact that someCaiophora clades are well supported, several infra-generic clades are poorly supported (Fig. 2). All phyloge-netic reconstructions recovered a clade, which includes the spe-cies Caiophora aconquijae Sleumer, Caiophora dumetorumUrb. & Gilg., C. clavata Urb. & Gilg. and C. lateritiaKlotzsch. C. arechavaletae was always recovered as alineage diverging early from the remaining Caiophoraclade (Fig. 2, Supplementary files 7–9: Figs. S2, S3and S4). Loasa was never recovered as a monophyleticgroup, and Loasa Ser. Pinnataewas always recovered as sisterto Caiophora (Fig. 2, Supplementary files 7–9: S2, S3 andS4). Reconstructions with the ITS region using differentmethods yielded remarkably different tree topologies and re-vealed that the majority of the groups within Caiophora arepoorly supported (Supplementary file 8: Fig. S3).Reconstructions with trnSGCU-trnGUUC resulted in relativelywell-supported groups within Caiophora, with Loasa ser.Pinnatae being always retrieved as sister to Caiophora(Supplementary file 9: Fig. S4).

According to the Bayesian inference (BI) analysisusing concatenated markers, Caiophora is recovered as amonophyletic and highly supported genus (Fig. 2;Table 1); species in Loasa ser. Pinnatae constitute a cladethat is sister to Caiophora. C. arechavaletae andC. nivalis Lillo are recovered as a lineage diverging earlyfrom the remaining Caiophora species. Most of the infra-generic groups proposed for Caiophora (Ackermann2012), which were defined on the basis of floral and veg-etative morphology, are not recovered as monophyletic(Fig. 2). The C. carduifolia and the C. cirsiifolia groups arenot monophyletic (Fig. 2), since C. carduifolia (1) constitutes

a relatively well-supported clade with C. coronata, andC. carduifolia (2) forms a relatively well-supported clade withC. chuquitensis. C. cirsiifolia (1) and C. cirsiifolia (2) consti-tute another relatively well-supported clade, which is sister toC. coronata-C. carduifolia (1). C. cirsiifolia (3) andC. cirsiifolia (4) also constitute a relatively well-supportedclade, sister to C. chuquitensis-C. carduifolia (2).Furthermore, C. coronata and C. pentlandii (both assignedto the C. coronata group) are not monophyletic (Fig. 2).Species in the C. lateritia and the C. clavata groups are notmonophyletic but together constitute a monophyletic group(Fig. 2). Only the C. pterosperma group (Caiophorastenocarpa Urb. & Gilg. and Caiophora dederichiorumMark. Ackermann & Weigend) is monophyletic and highlysupported (Fig. 2; Table 1).

Divergence times in Caiophora

According to the LRT, concatenated markers did not fit well to astrict clockmodel (lnL ratio test statistic (LRTS) = 79.93; df= 34;P < 0.01). The same result was obtained for non-concatenatedITS and trnSGCU-trnGUUC markers (LRTS = 343.55; df = 34;P < 0.01 and LRTS = 67.86; df = 34; P < 0.01, respectively).According to BI using relaxed clock, Caiophora originated16.05 Ma (95% highest posterior density (HPD) 7.02–26.12).The dating of the diversification of the genus using differentcalibration methods was estimated as follows: 10.43 Ma (95%HPD 4.37–17.64) (BI), 5.02 Ma (LRMD) and 15.17 Ma(NPRS). According to BI, most of the divergence events inCaiophora took place during the Late Pliocene and thePleistocene, i.e. during the last 5 million years (Fig. 2; Table 1).For BI estimation, effect sample sizes (ESSs) varied between 458and 2829.

Reconstruction of the evolution of pollination modes

Hummingbird pollination is not the ancestral condition withinCaiophora, since the species that diverged earlier from theremaining Caiophora clade are bee pollinated, and their re-constructed ancestors are likely to have been bee pollinated(Fig. 2). There is uncertainty regarding how many times hum-mingbird pollination evolved from bee pollination inCaiophora, since the common ancestor of the three lineagespresenting hummingbird pollinated species (C. carduifolia(1)-C. chuquitensis; C. hibiscifolia-Caiophora canarinoides-Caiophora madrequisa; C. carduifolia (2)-C. coronata) hadequal probabilities of being bee or hummingbird pollinated(Fig. 2). It is likely that there was a transition from humming-bird to rodent pollination inC. coronata (Fig. 2) and that therewas a reversion from hummingbird to bee pollination in thelineage giving rise to Caiophora cernua, C. clavata,C. dumetorum and C. lateritia (Fig. 2).

34 Strelin M.M. et al.

Discussion

Phylogenetic relationships

This study supports the monophyly of Caiophora and, in ac-cordance with previous studies (Hufford et al. 2003, 2005),supports the fact that Loasa ser. Pinnatae is the Loasa groupmore closely related to Caiophora (Fig. 2). Although someCaiophora clades are poorly supported, our study yielded abetter resolution of infra-generic clades than previous works(Weigend et al. 2004; Hufford et al. 2003, 2005). Low nodesupport values in several Caiophora clades may be a conse-quence of the reduced number of species and genetic markersused in this study, but it may also suggest an adaptive radiationin this genus, since the same macroevolutionary pattern wasrecovered for other Andean angiosperm lineages undergoingsuch diversification events (e.g. Smith and Baum 2006; Smithet al. 2008; Moré et al. 2015).

The infra-generic Caiophora groups, which were de-fined on the basis of floral and vegetative morphology(Ackermann 2012), were not recovered in the phylogeny,with the exception of the C. pterosperma group, which has ahigh support (Fig. 2; Table 1). Rather than reflecting phyloge-netic proximity, infra-generic groups in Caiophora may

reflect shared adaptations. Some of the traits used to defineinfra-generic groups in this genus, e.g. flower morphologyand colouration, as well as inflorescence architecture(Ackermann 2012), may have been under selection to someextent. For instance, variation in the morphology and thecolouration of the corolla is related to different functionalgroups of pollinators in the literature, e.g. Fenster et al.(2004). Further, the morphology and the colouration of thestaminodial complex, a flower structure unique to Loasoideaethat participates in nectar harvesting (Ackermann and Weigend2006) and in flower-pollinator mechanical fit (Weigend andGottschling 2006; Weigend et al. 2010), differ between beeand hummingbird pollinated species (Weigend andGottschling 2006). Indeed, Strelin et al. (2016a) demonstratedthat the shape of the corolla and the staminodial complex areunder pollinator selection in Caiophora. Interestingly, whilespecies in theC. clavata and theC. cirsiifolia groups are mainlypollinated by bees, hummingbird pollination appears to be im-portant for species in the C. lateritia and the C. carduifoliagroups (Fig. 2; Supplementary file 3: Table S3). Concluding,some of the trait combinations used in the previous delimitationof Caiophora infra-generic groups may reflect adaptations todifferent functional groups of pollinators rather than shared an-cestry. Nevertheless, a lack of correspondence between plant

#

*

#

#

*

*

*

#

*

+

+

$

$

&

&

&

¶

£

£

¢

§

§

¥

¤

~

* Lateritia

# Clavata

$ Pterosperma

+ Coronata

Bees

Hummingbirds

Small rodents

Unknown pollinator

& Pinnatae

~ Pusillae

¶ Volubiles

¥ Floribundae

¤ Loasa

§ Macrospermae

£ Gripidea

¢ Angulatae

Assigned group

Pollinator

Assigned group

and pollinator

₣

€

◊

Ω

₣

₣

₣

₣ Cirsiifolia

Ω

Ω Carduifolia

Arechavaletae

◊ Nivalis

€ Chuquitensis

Puna-Altiplano

Plateau

Cordillera Occidental

Cordillera Central

Cordillera Oriental (arid)

Cordillera Oriental (yungas)

Atacama dessert

Sierras Subandinas

Sierras Pampeanas

Precordillera

Cordillera Frontal

Cordillera Principal

Patagonian Andes

Pampas grasslands

Chilean scrub

Central Brazilian sabanas

Caiophora

Loasa

Blumenbachia

Aosa

Xylopodia

*

C. dumetorum

C. lateritia

C. clavata

C. cernua

C. madrequisa

C. buraeavii

C. canarinoides

C. chuquisacana

C. hibiscifolia

C. cirsiifolia (2)

C. cirsiifolia (1)

C. carduifolia (1)

C. coronata

C. carduifolia (2)

C. chuquitensis

C. cirsiifolia (3)

C. cirsiifolia (4)

C. pentlandii

C. dederichiorum

C. stenocarpa

C. arechavaletae

C. nivalis

L. bergii (1)

L. bergii (2)

L. pinnatifida

L. martini

B. insignis

B. amana

B. silvestris

L. heterophylla

L. acerifolia

L. pallida

X. klaprothioides

C. aconquijae

L. acanthifolia

A. rostrata

r

nt

o

p

7

Myr ago10203040

Paleogene

Oligocene Mio

Neogene Qua

Plio Plei

Cenozoic

6

5

4

2

1

8

10

9

11

12

15

16

17

18

19

20

23

24

25

26

27

28

31

32

33

Eocene

050

16.05

13

10.43

22

5.64

14

5.98

3

21

34

29

30

4

19

2

% Klaprothieae

%

Loasa

Fig. 2 Phylogenetic relationships, geographic distribution andpollination modes of Caiophora (Loasaceae, subfam. Loasoideae) andallied genera (Loasa and Blumenbachia). Geographic distributioncorresponds to the information in Supplementary file 2: Table S2. Thetopology of the phylogeny is the maximum posterior credibility tree of aBayesian analysis performed with ITS1, ITS2 and trnSGCU-trnGUUC

markers. Node support and dating values in Caiophora and basalgenera can be consulted in Table 1, using the reference number on eachnode. A reconstruction of ancestral pollination modes using stochasticcharacter mapping is shown. Pie charts on each node indicate theposterior probability of each pollinator, retrieved by 1000 stochasticcharacter mappings

Diversification of Caiophora (Loasaceae subfam. Loasoideae) 35

phenotype and phylogeny may, at least in part, also be a con-sequence of interspecific hybridization, which was already re-ported for some Caiophora species (Ackermann et al. 2008).Due to the low sample size of Caiophora species in this study,we are not able to arrive at taxonomic conclusions concerningthe different positions of the C. cirsiifolia and C. carduifoliaaccessions in the phylogeny. Nevertheless, since these specieswere previously defined based only on plant morphology(Ackermann 2012), it would not be surprising that

C. cirsiifolia (1)-C. cirsiifolia (2), C. cirsiifolia (3)-C. cirsiifolia (4), C. carduifolia (1) and C. carduifolia (2) rep-resent lineages with independent evolutionary trajectories.

Timing and context of Caiophora diversification

Three types of methods to estimate divergence times areoften recognized: methods that assume a global rate,methods that assume local rates and methods that accountfor rate heterogeneity through the phylogeny. Because ourmolecular clock analysis showed that there is no globalrate, we used methods of local rates (LRMD) and rate het-erogeneity. Among the latter, we used two differentmethods, (1) NPRS, a method that models the rate accord-ing to a standard Poisson process and the phylogeneticautocorrelation of rates and (2) the Bayesian relaxed clockapproach implemented in BEAST. The differences that wefound in the divergence time estimation with these threemethods are attributable to the differences in their assump-tions and statistical implementations (see Rutschmann2006, for a more detailed discussion on methods ofdivergence time estimations).

The origin and diversification of Caiophora can be re-lated to a variety of environmental events taking placesince the Early Miocene (Table 2). Our results suggest thatCaiophora originated during the Early-Middle Miocene,approximately 16 Ma. This period coincides with the de-formation and uplift of the Cordillera Frontal (∼17 to14 Ma) (Levina et al. 2014) and the Cordillera Principal(∼20 to 9 Ma) (Ramos et al. 2002; Giambiagi and Ramos2002). Interestingly, while the distribution of Caiophoraencompasses these two mountain ranges, which are in turnthe southern limit of the distribution of this genus (Fig. 2;Supplementary file 2: Table S2), species in Loasa ser.Pinnatae (sister to Caiophora in our phylogeny), as wellas the majority of the sampled species in clades that aresuccessive sister groups to Caiophora, only grow to thesouth of the Cordillera Frontal and the CordilleraPrincipal (Fig. 2; Supplementary file 2: Table S2).Moreover, Scyphanthus Sweet (not included in this study),which according to Hufford et al. (2003, 2005) is the genusmost closely related to Caiophora, also grows to the southof these mountain ranges (Ackermann 2012). This suggeststhat Caiophora may have originated as a consequence oflineage isolation in northern and higher-elevation areas,during the uplift of the Cordillera Frontal and theCordillera Principal. Pacific marine transgressions affect-ing central and northern Patagonia during the Early andMiddle Miocene (∼23–16 Ma) (Bechis et al. 2014) mayhave also fostered this split (Fig. 2; Supplementary file 3:Table S3).

According to our estimations, Caiophora began to diversi-fy about 5–15 Ma, depending on the dating method. If the

Table 1 Node support and dating values in Caiophora (Loasaceae,subfamily Loasoideae) and allied genera

Node MP/NJ/ML/BI support Age 95% HPD

1 – 44.68 –

2 –/–/99.6/0.99 26.97 13.74–41.00

3 –/–/–/– 20.76 9.006–35.39

4 –/–/67.8/0.86 16.97 5.97–29.25

5 100/100/100/1 5.86 0.91–12.51

6 98.9/–/–/0.75 23.81 10.33–36.82

7 98.97/100/99.9/1 11.83 3.29–22.18

8 100/100/100/1 3.72 0.60–8.53

9 74.93/68.59/98.5/1 20.09 8.27–31.97

10 100/100/100/0.99 16.05 7.02–26.12

11 100/100/99.7/1 4.81 0.59–10.71

12 94.62/100/98.6/0.97 2.12 0.20–5.38

13 58.53/56.09/71.4/1 10.43 4.37–17.64

14 –/–/–/0.76 5.98 1.11–11.99

15 71.5/–/–/0.61 8.43 3.36–14.35

16 88.82/87.9/96/0.99 4.03 0.53–8.70

17 –/53.59/–/0.61 6.74 2.58–11.63

18 –/50.29/–/0.74 3.85 3.36–14.35

19 –/–/–/– 1.98 0.13–5.06

20 –/–/66.3/0.86 0.67 0.05–2.2

21 –/–/51.1/0.98 1.77 0.22–4.21

22 –/–/–/0.61 5.64 2.12–9.77

23 –/–/–/0.83 3.18 0.55–6.37

24 –/–/–/0.87 1.66 0.19–3.90

25 –/60.79/–/0.83 1.30 0.09–3.48

26 –/–/–/0.53 4.65 1.69–8.21

27 –/–/–/0.76 2.51 0.38–5.34

28 –/–/–/0.53 3.58 1.22–6.53

29 –/–/–/0.53 2.76 1.22–6.53

30 –/–/–/– 1.72 1.22–6.53

31 –/70.4/50.6/0.53 1.83 0.48–3.62

32 86.1/70.4/50.06/0.99 1.14 0.22–2.46

33 96.31/–/87/1 0.34 0.03–0.94

34 –/–/–/– 0.15 0.03–0.45

Caiophora group values are in italics. Only node support values >50%/0.5 are shown

MP maximum parsimony, NJ neighbour joining, ML maximum likeli-hood, BI Bayesian inference

36 Strelin M.M. et al.

NPRS estimation (∼15 Ma) and the BI estimation (∼10 Ma)are considered, the diversification of Andean hummingbirds(∼15 Ma) (McGuire et al. 2014) may have contributed to the

diversification of Caiophora. The reliance of many species inthe genus on hummingbird pollination, as well as the highposterior probability of a relatively early acquisition of

Table 2 Summary of evolutionary events, i.e. origin and divergence of different groups, inCaiophora and allied genera, along with the environmentalevents hypothesized to be involved in these evolutionary events

Evolutionary event Explanatory environmental event Geological epoch(environmental event)

Origin of Caiophora (16.05 Ma [7.02–26.12]) (BI) Main uplift event of the Cordillera Principal (∼20 to 9Ma; Ramos et al. 2002 and Giambiagi and Ramos 2002).Uplift event of the Cordillera Frontal (∼17 to 14 Ma;Levina et al. 2014).

Early to Late Miocene

Origin of Caiophora (16.05 Ma [7.02–26.12]) (BI) Pacific marine transgressions in central and northernPatagonia (∼23 to 16 Ma; Bechis et al. 2014)

Early to Middle Miocene

Diversification of Caiophora (10.43 Ma [4.37–17.64])(BI); 5.02 Ma (LRMD); 15.17 Ma (NPRS)

Diversification of the Andean hummingbird lineage(∼15 Ma; McGuire et al. 2014).

Middle Miocene

Diversification of Caiophora (10.43 Ma [4.37–17.64])(BI); 5.02 Ma (LRMD); 15.17 Ma (NPRS)

Uplift of the Cordillera Oriental and the Sierras Subandinas(eastern Andean slope). The Puna-Altiplano Plateaurose from 1500 to 4000 m. (∼10 Ma to present:Armijo et al. 2015).

Middle Miocene to present

Diversification of Caiophora (10.43 Ma [4.37–17.64])(BI); 5.02 Ma (LRMD); 15.17 Ma (NPRS)

The Puna-Altiplano Plateau rose ∼2.5 km (∼10 to 7 Ma;Garzione et al. 2006; Hoke and Garzione 2008).The Puna-Altiplano Plateau rose from 1500 to 2500m a.s.l (∼9 to 5 Ma: Lamb et al. 1997). Uplift of theSierras Subandinas (east to the Puna-Altiplano Plateau)and the Cordillera Oriental. (∼10 to 6 Ma; Hoke andGarzione 2008). This uplift propagates into the SierrasSubandinas (∼9 Ma to present; Echavarria et al. 2003)

Late Miocene to present

Origin of Caiophora clade composed of speciesinhabiting vegetated areas (towards the east) of thesub-Andean mountain range (C. clavata, C. cernua,C lateritia, C. dumetorum, C. aconquijae), and theCordillera Oriental (C. hibiscifolia, C. madrequisa,C. canarinoides, C. buraeavii, C. chuquisacana)(5.64 Ma [2.12–9.77]) (BI)

Uplift of the Cordillera Oriental and establishment ofthe mountain rainforest (Yungas; ∼7 to 6 Ma; Grahamet al. 2001). Establishment of the Yungas on the easternslope of the Sierras Subandinas (∼4 Ma; Starck andAnzótegui 2001).

Late Miocene to Pliocene

Origin and diversification of the Caiophora cladecomposed of species inhabiting the SierrasSubandinas and the Sierras Pampeanas(C. clavata, C. cernua, C. lateritia, C. dumetorum,C. aconquijae) (2.76 Ma [1.22–6.53] and 1.83Ma [0.48–3.62], respectively (BI)

Uplift and deformation of the Sierras Subandinas on theBrasilia Massif (∼5 to 0 Ma; Lamb et al. 1997).

Pliocene to present

Origin of the C. carduifolia (1)-C. coronata clade(3.18 Ma [0.55–6.37]) (BI)

The Puna-Altiplano Plateau rose from 2500 to 4000m a.s.l. (∼5 to Ma to present; Lamb et al. 1997; Barnesand Ehlers 2009)

Pliocene to present

Origin of the C. carduifolia (2)-C. chuquitensisclade (3.85 Ma [3.36–14.35]) (BI)

The Puna-Altiplano Plateau rose from 2500 to 4000m.a.s.l. (∼5 Ma to present; Lamb et al. 1997; Barnesand Ehlers 2009)

Pliocene to present

Origin of the C. cirsiifolia (1)-C. cirsiifolia (2) andC. cirsiifolia (3)-C. cirsiifolia (4) clades on thewestern Andean slope (3.85 Ma [3.36–14.35]and 3.18 Ma [0.55–6.37], respectively) (BI)

The Puna-Altiplano Plateau rose from 2500 to 4000m.a.s.l. (∼5 to Ma to present; Lamb et al. 1997; Barnesand Ehlers 2009)

Pliocene to present

Split between the lineage of Andean Caiophoraspecies and C. arechavaletae (inhabiting thePampas grasslands) (5.98 Ma [1.11–11.99]) (BI)

Small dust particles are transported by dust storms formthe Puna-Altiplano Plateau (NW of Argentina) to thePampas grasslands and deposited by rain (∼3 Ma topresent; Gaiero et al. 2013)

Late Pliocene to present

Origin of the lineage giving rise to B. insignisand B. amana (inhabiting the Pampas grasslandsand the Central Brazilian sabanas) (11.83 Ma[3.29–22.18]) (BI)

Paranense marine transgression (∼13 to 15 Ma; Donatoet al. 2003)

Middle Miocene

Diversification of Caiophora (Loasaceae subfam. Loasoideae) 37

hummingbird pollination in Caiophora (Fig. 1), is in accor-dance with this expectation. Our estimation using BI suggeststhat most of the divergence events within Caiophora are rel-atively recent (<5 Ma). The fact that several species in thegenus hybridize giving rise to fertile descendants(Ackermann et al. 2008) and that important rearrangementsof flower development, which are expected in adaptive radia-tion context (West-Eberhard 2005), apparently took place inCaiophora (Strelin et al. 2016b) is consistent with a recentdiversification. Interestingly, the Late Miocene, the Plioceneand the Pleistocene were periods of intense orogenic activityin the Central Andes and related areas (Armijo et al. 2015).The Puna-Altiplano Plateau rose ∼2500 m. during the LateMiocene (∼10 to 7 Ma) (Garzione et al. 2006; Hoke andGarzione 2008), and the uplift of the Cordillera Oriental (eastto the Puna-Altiplano Plateau) also took place during this pe-riod (∼10 to 6 Ma) (Hoke and Garzione 2008). This upliftcontinued eastwards, into the Sierras Subandinas (∼5 Ma topresent, according to Lamb et al. (1997); ∼9 Ma to presentaccording to Echavarria et al. (2003)). Interestingly, althougheastern AndeanCaiophora clades are poorly supported, diver-gence within this genus also seems to follow this west to eastdirection (Fig. 2).

During the Miocene and the Pliocene, the rising CordilleraOriental and the Sierras Subandinas began to act as a rainbarrier, retaining and precipitating Atlantic humidity. Thisled to the establishment of the mountain rain forests(Yungas) on the eastern slopes of both mountain ranges about7 to 4 Ma (Graham et al. 2001; Starck and Anzótegui 2001).The origin of the clade composed of C. cernua (Griseb.) Urb.& Gilg ex Kurtz, C. hibiscifolia (Griseb.) Urb. & Gilg,C. madrequisa Killip, C. canarinoides (Lenné & C. Koch)Urb. & Gilg, Caiophora buraeavii Urb. & Gilg, Caiophorachuquisacana Urb. & Gilg., C. clavata, C. lateritia,C. dumetorum and C. aconquijae, which are mainly recordedfor rain forest areas of the Sierras Subandinas and theCordillera Oriental (Fig. 2; Supplementary file 2: Table S2)is dated approximately 5 Ma. In turn, the origin and diversifi-cation of the Caiophora subclade composed of C. clavata,C. cernua, C. lateritia, C. dumetorum and C. aconquijae,some of which are recorded for the Cordillera Oriental, butwhich are also recorded for mountain ranges to the east(Sierras Subandinas and Sierras Pampeanas) (Fig. 2;Supplementary file 2: Table S2), took place ∼3 and ∼2 Ma,respectively (Table 1). Although these clades are poorly sup-ported, this divergence pattern is consistent with the eastwardpropagation of Andean uplift (Table 2). The remainingCaiophora species in this study were only recorded in aridenvironments, to the west of the distribution range of the pre-viously mentioned species (Fig. 2; Supplementary file 2:Table S2). Reconstruction of ancestral pollination modes sug-gests a reversion from hummingbird to bee pollination in theCaiophora lineage inhabiting the Sierras Subandinas

(C. dumetorum, C. lateritia, C. cernua and C. clavata).Since the Cordillera Oriental (towards the west) is higher thanthe Sierras Subandinas (towards the east) and insect pollina-tion is more frequent at lower-elevation areas (Arroyo et al.1985), reversion from hummingbird to bee pollination in thiseastern Caiophora lineage may be expected.

Orogenic activity during the Pliocene-Pleistocene did af-fect not only the eastern slope of the Andes, but also the Puna-Altiplano Plateau, which rose from 2.500 m. to the current4.000 m. (Lamb et al. 1997; Barnes and Ehlers 2009). Thissudden uplift of the Puna-Altiplano Plateau presumably drovediversification in Caiophora through the generation of high-elevation pollination niches, where insects are scarce and re-liance on vertebrate pollinators may therefore be important,e.g. Perez et al. (2006). Interestingly, two small and well-supported high-Andean Caiophora clades (2000 to5000 m.), C. coronata-C. carduifolia (1) and C. chuquitensisandC. carduifolia (2) (Fig. 2; Supplementary file 2: Table S2),seem to have originated during the end of the Pliocene and tohave diverged during the Pleistocene. Caiophora coronata-C. carduifolia (1) originated approximately 3.18 Ma and di-verged approximately 1.66Ma;C. chuquitensis-C. carduifolia(2) originated approximately 3.85 Ma and diversified approx-imately 1.77 Ma. Although bee pollination was reported forsome of the species in these clades, all of them also rely onvertebrate pollinators. Hummingbirds play an important rolein the pollination of C. chuquitensis and C. carduifolia(Ackermann and Weigend 2006; Fig. 2; Supplementary file3: Table S3), and small rodents are in turn pollinators ofC. coronata (Cocucc i and Sérs ic 1998; Fig . 2 ;Supplementary file 3: Table S3). While the Puna-AltiplanoPlateau uplift may have led to adaptation of high-AndeanCaiophora species to hummingbird and rodent pollination,this same event may have led to the geographic isolation ofthe C. cirsiifolia (1)-C. cirsiifolia (2) and the C. cirsiifolia (3)-C. cirsiifolia (4) clades on the western Andean slope, morespecifically on the Cordillera Occidental (Fig. 2;Supplementary file 2: Table S2). Reconstructions of ancestralpollination modes are ambiguous with regard to whether theacquisition of hummingbird pollination evolved or not inde-pendently in the C. coronata-C. carduifolia (1) and theC. chuquitensis-C. carduifolia (2) lineages.

The origin of C. arechavaletae deserves an explanationalternative to Andean orogeny.WhileC. arechavaletae is onlydistributed in the Uruguayan and the Brazilian Pampas grass-lands (Fig. 2; Supplementary file 2; Table S2), all remainingCaiophora species are associated with the Andes (Fig. 2;Supplementary file 2; Table S2). The split between theAndean Caiophora lineage and C. arechavaletae is datedabout 6 Ma (Fig. 2). Neither orogeny nor climate change canaccount for this split (Jorge Strelin, personal communication),but aeolian transportation of Caiophora seeds from theAltiplano region to the current Pampas grasslands is a

38 Strelin M.M. et al.

plausible explanation. Gaiero et al. (2013) described two re-cent events of major dust transportation originating in thehigh-altitude, subtropical Puna-Altiplano Plateau (15–26° S;65–69° W). According to paleoenvironmental evidence, theseevents may have also taken place during the Late Pliocene,about 5 Ma. Large dust particles from these events were de-tected in traps in the north of the province of Buenos Aires,and according to the modelling of particle deposition and con-centration, these dust storms may have also transported largedust particles from the Puna-Altiplano Plateau to theUruguayan and Brazilian grasslands (Gaiero et al. 2013).Interestingly, the seeds of most of the BSouth AndeanLoasas^, which include Loasa, Caiophora, Scyphanthus andBlumenbachia, are dispersed by wind (Weigend et al. 2005).Seeds present specialized testa structures that constitute anadaptation for aeolic transportation (Weigend et al. 2005).Based on the weight, the linear measurements and the wing-like structures of C. arechavaletae seeds (Weigend et al.2005), strong winds may have transported Caiophora seedsfrom the Puna-Altiplano Plateau to the Pampas grasslands,where these may have been deposited by rain (Gaiero, person-al communication).

Although the sampling effort in this work concentrated onCaiophora and not on its allied lineages, it can be speculatedthat the origin of the Blumenbachia insignis Schrad.-Blumenbachia amana T. Henning & Weigend clade (about12 Ma), which are recorded for the lowland and grow closerto the Atlantic coast than to the Andes (Fig. 2; Supplementaryfile 2; Table S2), is related to the Paranaense marine transgres-sion that took place during the Middle-Late Miocene (15–13Ma), isolating the current Pampas grasslands from the westof southern South America (Donato et al. 2003) (Fig. 2).

Concluding, this study suggests that Andean orogeny fos-tered the origin and diversification of Caiophora, through thegeneration of vicariant barriers and novel ecological condi-tions, e.g. emergence of mountain rain forests and dependenceon vertebrate pollination in high-Andean environments.Unresolved infra-generic phylogenetic relationships, alongwith evidence that many divergence events inCaiophora havetaken place during the last 5 Myr, suggest a rapid diversifica-tion of the genus. Nevertheless, phylogeographic studies,along with work looking more closely at ecological factorspromoting lineage divergence in Caiophora, e.g. plant-pollinator interactions and seed dispersion, are essential tobetter understand the overwhelming diversification of this ge-nus. Including more Caiophora species, along with species inits sister group, as well as using more genetic markers, is alsoan essential step to accomplish this goal.

Acknowledgements We thank Marcela Moré and Cristina Acosta(IMBIV, CONICET, Universidad Nacional de Córdoba, Argentina);Romina Vidal-Russel (INIBIOMA, CONICET, Universidad Nacionaldel Comahue); Jorge Strelin, Mateo Martini and Diego Gaiero

(CICTERRA, Universidad Nacional de Córdoba, Instituto AntárticoArgentino); Matías Ghiglione (Instituto de Estudios Andinos,CONICET, Universidad Nacional de Buenos Aires, Argentina); and thetwo anonymous reviewers for contributing with their comments and sug-gestions to the quality of this manuscript. We thank Laura Gatica forediting the English of this manuscript and Maximilian Weigend (NeesInstitut für Biodiversität der Pflanzen, Rheinische Friedrich-Wilhelms-Universität, Bonn, Germany) for providing material, funds and data forthis study. M.S. has a scholarship from the National Scientific andTechnical Research Council (CONICET).

References

Ackermann, M. (2012). Studies on systematics, morphology and taxon-omy of Caiophora and reproductive biology of Loasaceae andMimulus (Phrymaceae). PhD Thesis. Free University of Berlin,Germany.

Ackermann, M., & Weigend, M. (2006). Nectar, floral morphology andpollination syndrome in Loasaceae subfam. Loasoideae (Cornales).Annals of Botany, 98, 503–514.

Ackermann, M., Achatz, M., & Weigend, M. (2008). Hybridization andcrossability in Caiophora (Loasaceae, subfam. Loasoideae): are in-terfertile species and inbred populations results of a recent radiation?American Journal of Botany, 95, 1109–1121.

Altshuler, D. L., Dudley, R., & McGuire, J. A. (2004). Resolution of aparadox: hummingbird flight at high elevation does not come with-out a cost. Proceedings of the National Academy of Sciences of theUnited States of America, 101, 17731–17736.

Antonelli, A., & Sanmartín, I. (2011). Why are there so many plantspecies in the Neotropics? Taxon, 60, 403–414.

Antonelli, A., Nylander, J. A., Persson, C., & Sanmartín, I. (2009).Tracing the impact of the Andean uplift on Neotropical plant evolu-tion. Proceedings of the National Academy of Sciences, 106, 9749–9754.

Armijo, R., Lacassin, R., Coudurier-Curveur, A., & Carrizo, D. (2015).Coupled tectonic evolution of Andean orogeny and global climate.Earth-Science Reviews, 143, 1–35.

Arroyo, M. T. K., Primack, R., & Armesto, J. J. (1985). Communitystudies in pollination ecology in the high temperate Andes ofCentral Chile. I. Pollination mechanism and altitudinal variation.American Journal of Botany, 69, 82–97.

Baranzelli, M. C., Johnson, L. J., Cosacov, A., & Sérsic, A. N.(2014). Historical and ecological divergence among popula-tions of Monttea chilensis (Plantaginaceae), an endemic endan-gered shrub bordering the Atacama Desert, Chile. EvolutionaryEcology, 28, 751–774.

Barnes, J. B., & Ehlers, T. A. (2009). End member models for AndeanPlateau uplift. Earth-Science Reviews, 97, 105–132.

Bechis, F., Encinas, A., Concheyro, A., Litvak, V. D., Aguirre-Urreta, B.,& Ramos, V. A. (2014). New age constraints for the Cenozoic ma-rine transgressions of northwestern Patagonia, Argentina (41°-43°S): paleogeographic and tectonic implications. Journal of SouthAmerican Earth Sciences, 52, 72–93.

Bryson Jr., R. W., García-Vázquez, U. O., & Riddle, B. R. (2012).Relative roles of Neogene vicariance andQuaternary climate changeon the historical diversification of bunchgrass lizards (Sceloporusscalaris group) in Mexico.Molecular Phylogenetics and Evolution,62, 447–457.

Cavides-Vidal, E., Bozinovic, F., & Rosenmann, M. (1987). Thermalfreedom of Graomys griseoflavus in a seasonal environment.Comparative Biochemistry and Physiology Part A: Physiology, 87,257–259.

Diversification of Caiophora (Loasaceae subfam. Loasoideae) 39

Chaves, J. A., Weir, J. T., & Smith, T. B. (2011). Diversification inAdelomyia hummingbirds follows Andean uplift. MolecularEcology, 20, 4564–4576.

Clapperton, C. H. (1983). The glaciation of the Andes. QuaternaryScience Reviews, 2, 83–155.

Cocucci, A. A., & Sérsic, A. N. (1998). Evidence of rodent pollination inCajophora coronata (Loasaceae). Plant Systematics and Evolution,211, 113–128.

Cosacov, A., Sérsic, A. N., & Sosa, V. (2010). Multiple periglacial refugiain the Patagonian steppe and post-glacial colonization of the Andes:the phylogeography of Calceolaria polyrhiza. Journal ofBiogeography, 37, 1463–1477.

Cruden, R. W. (1972). Pollinators in high-elevation ecosystems: relativeeffectiveness of birds and bees. Science, 176, 1439–1440.

Donato, M., Posadas, P., Miranda-Esquivel, D. R., Jaureguizar, E. O., &Cladera, G. (2003). Historical biogeography of the Andean region:evidence from Listroderina (Coleoptera: Curculionidae:Rhytirrhinini) in the context of the South American geobiotic sce-nario. Biological Journal of the Linnean Society, 80, 339–352.

Drummond, C. S., Eastwood, R. J., Miotto, S. T. S., & Hughes, C. E.(2012a). Multiple continental radiations and correlates of diversifi-cation in Lupinus (Leguminosae): testing for key innovations withincomplete taxon sampling. Systematic Biology, 61, 443–460.

Drummond, A. J., Suchard, M. A., Xie, D., & Rambaut, A. (2012b).Bayesian phylogenetics with BEAUti and the BEAST 1.7.Molecular Biology and Evolution, 29, 1969–1973.

Echavarria, L., Hernndez, R., Allmendinger, R., & Reynolds, J. (2003).Subandean thrust and fold belt of northwestern Argentina: geometryand timing of the Andean evolution. AAPG Bulletin, 87, 965–985.

Felsenstein, J. (1984). Distance methods for inferring phylogenies: a jus-tification. Evolution, 38, 16–24.

Fenster, C. B., Armbruster, W. S., Wilson, P., Dudash, M. R., &Thomson, J. D. (2004). Pollination syndrome and floral specializa-tion. Annual Review of Ecology, Evolution and Systematics, 25,375–403.

Gaiero, D. M., Simonella, L., Gassó, S., Gili, S., Stein, A. F., Sosa, P.,Becchio, R., Arce, J., & Marelli, H. (2013). Ground/satellite obser-vations and atmospheric modeling of dust storms originating in thehigh Puna-Altiplano deserts (South America): implications for theinterpretation of paleo-climatic archives. Journal of GeophysicalResearch: Atmospheres, 118, 3817–3831.

Gamble, T., Simons, A. M., Colli, G. R., & Vitt, L. J. (2008). Tertiaryclimate change and the diversification of the Amazonian gecko ge-nus Gonatodes (Sphaerodactylidae, Squamata). MolecularPhylogenetics and Evolution, 46, 269–277.

Garzione, C. N., Molnar, P., Libarkin, J. C., & MacFadden, B. J. (2006).Rapid late Miocene rise of the Bolivian Altiplano. Evidence fromremoval of mantle lithosphere. Earth and Planetary Science Letters,241, 543–556.

Giambiagi, L. B., & Ramos, V. A. (2002). Structural evolution of theAndes in a transitional zone between flat and normal subduction(33°30′–33°45′S), Argentina and Chile. Journal of SouthAmerican Earth Studies, 15, 101–116.

Graham, A., Gregory-Wodzicki, K. M., & Wright, K. L. (2001). Studiesin Neotropical Paleobotany. XV. A Mio-Pliocene palynoflora fromthe Eastern Cordillera, Bolivia: implications for the uplift history ofthe Central Andes. American Journal of Botany, 88, 1545–1557.

Hamilton, M. B. (1999). Four primer pairs for the amplification of chlo-roplast intergenic regions with intraspecific variation. MolecularEcology, 8, 521–523.

Hoiss, B., Krauss, J., Potts, S. G., Roberts, S., & Steffan-Dewenter, I.(2012). Altitude acts as an environmental filter on phylogeneticcomposition, traits and diversity in bee communities. Proceedingsof the Royal Society B-Biological Sciences, 279, 4447–4456.

Hoke, G. D., & Garzione, C. N. (2008). Paleosurfaces, paleoelevation,and the mechanisms for the late Miocene topographic development

of the Altiplano plateau. Earth and Planetary Science Letters, 271,192–201.

Hufford, L., McMahon, M. M., Sherwood, A. M., Reeves, G., & Chase,M.W. (2003). The major clades of Loasaceae: phylogenetic analysisusing the plastid matK and trnL-trnF regions. American Journal ofBotany, 90, 1215–1228.

Hufford, L., McMahon, M. M., O’Quinn, R., & Poston, M. E. (2005). Aphylogenetic analysis of Loasaceae, subfamily Loasoideae based onplastid DNA sequences. International Journal of Plant Sciences,166, 289–300.

Hughes, C., & Eastwood, R. (2006). Island radiation on a continentalscale: exceptional rates of plant diversification after uplift of theAndes. Proceedings of the National Academy of Sciences of theUnited States of America, 103, 10334–10339.

Jobb, G., Haeseler, A. V., & Strimmer, K. (2004). TREEFINDER: apowerful graphical analysis environment for molecular phyloge-netics. BMC Evolutionary Biology. doi:10.1186/1471-2148-4-18.

Katoh, K., & Standley, D. M. (2013). MAFFT multiple sequence align-ment software version 7: improvements in performance and usabil-ity. Molecular Biology and Evolution, 30, 772–780.

Lamb, S., Hoke, L., Kennan, L., & Dewey, J. (1997). Cenozoic evolutionof the Central Andes in Bolivia and northern Chile. GeologicalSociety, London, Special Publications, 121, 237–264.

Levina, M., Horton, B. K., Fuentes, F., & Stockli, D. F. (2014). Cenozoicsedimentation and exhumation of the foreland basin system pre-served in the Precordillera thrust belt (31–32°S), southern centralAndes, Argentina. Tectonics, 33, 1659–1680.

Luebert, F., & Weigend, M. (2014). Phylogenetic insights into Andeanplant diversification. Frontiers in Ecology and Evolution.doi:10.3389/fevo.2014.00027.

Martinez, J. J., Ferro, L. I., Mollerach, L. I., & Barquez, R. M. (2012).The phylogenetic relationships of the Andean swamp rat genusNeotomys (Rodentia, Cricetidae, Sigmodontinae) based on mito-chondrial and nuclear markers. Acta Theriologica, 57, 277–287.

Martini, M.A., Strelin, J.A., Kaplan,M.R., & Schaefer, J.M. (2012). Glacialand periglacial geomorphology and chronology around theNevado deChañi (Cordillera Oriental of Jujuy): implication for past climate inNW Argentina. Resource document. AGU Fall Meeting Abstracts.http://adsabs.harvard.edu/abs/2012AGUFMEP53C1053M. Accessed13 Oct 2016.

McGuire, J. A., Witt, C. C., Remsen Jr., J. V., Corl, A., Daniel, L.,Rabosky, D. L., Altshuler, D. L., & Dudley, R. (2014). Molecularphylogenetics and the diversification of hummingbirds. CurrentBiology, 24, 910–916.

Meng, H. H., & Zhang, M. L. (2013). Diversification of plant species inarid Northwest China: species-level phylogeographical history ofLagochilus Bunge ex Bentham (Lamiaceae). MolecularPhylogenetics and Evolution, 68, 398–409.

Mercer, J. H., & Sutter, J. (1982). Late Miocene—earliest Pliocene gla-ciation in Southern Argentina: implications for global ice-sheet his-tory. Palaeogeography, Palaeoclimatology, Palaeoecology, 38,185–206.

Minelli, A., & Fusco, G. (2012). On the evolutionary developmentalbiology of speciation. Evolutionary Biology, 39, 242–254.

Moré, M., Cocucci, A. A., & Sérsic, A. N. (2015). Phylogeny and floraltrait evolution in Jaborosa (Solanaceae). Taxon, 64, 523–534.

Muschick, M., Indermaur, A., & Salzburger, W. (2012). Convergent evo-lution within an adaptive radiation of cichlid fishes.Current Biology,22, 2362–2368.

Mutke, J., Jacobs, R., Meyers, K., Henning, T., & Weigend, M. (2014).Diversity patterns of selected Andean plant groups correspond totopography and habitat dynamics, not orogeny. Frontiers inGenetics. doi:10.3389/fgene.2014.00351.

Müller, J., Müller, K., Quandt, D., & Neinhuis, C. (2010). PhyDe-Phylogenetic Data Editor. Program distributed by the author.Available at: http://www.phyde.de/index.html.

40 Strelin M.M. et al.

Nespolo, R. F., Opazo, J. C., & Rosenmann, F. B. (1999). Thermal accli-mation, maximum metabolic rate, and nonshivering thermogenesisof Phyllotis xanthopygus (Rodentia) in the Andes Mountains.Journal of Mammalogy, 80, 742–748.

Nielsen, R. (2002). Mapping mutations on phylogenies. SystematicBiology, 51, 729–739.

Nylin, S., Slove, J., & Janz, N. (2014). Host plant utilization, host rangeoscillations and diversification in Nymphalid butterflies: a phyloge-netic investigation. Evolution, 68, 105–124.

Palazzesi, L., Gottschling, M., Barreda, V., & Weigend, M. (2012). FirstMiocene fossils of Vivianiaceae shed new light on the phylogeny,divergence times, and historical biogeography of Geraniales.Botanical Journal of the Linnean Society, 107, 67–85.

Perez, F., Arroyo, M. T. K., Medel, R., & Hershkovitz, M. A. (2006).Ancestral reconstruction of flower morphology and pollination sys-tems in Schizanthus (Solanaceae). American Journal of Botany, 93,1029–1038.

Ramos, V. A., Cristallini, E. O., & Pérez, D. J. (2002). The Pampean flat-slab of the Central Andes. Journal of South American Earth Studies,15, 59–78.

Revell, L. J. (2012). Phytools: an R package for phylogenetic compara-tive biology (and other things). Methods in Ecology and Evolution,3, 217–223.

Revell, L. J. (2014). Graphical methods for visualizing comparative dataon phylogenies. In L. Z. Garamszegi (Ed.), Modern phylogeneticcomparative methods and their application in evolutionary biology(pp. 77–103). Berlin, Heidelberg: Springer.

Ronquist, F., & Huelsenbeck, J. P. (2003). MrBayes 3: Bayesian phylo-genetic inference under mixed models. Bioinformatics, 19, 1572–1574.

Rubioldo, D., Seggiaro, R., Gallardo, E., Disalvo, A., Sanchez,M., Turel,A., Ramallo, E., Sandruss, A., & Godeas, M. (2001). HojaGeológica 2366-II / 2166-IV, La Quiaca. Geología y Provincias deJujuy y Salta. Instituto de Recursos Minerales, Servicio GeológicoMinero Argentino. Boletín 246, p. Buenos Aires.

Rutschmann, F. (2006). Molecular dating of phylogenetic trees: a briefreview of current methods that estimate divergence times. Diversityand Distributions, 12, 35–48.

Sanderson, M. J. (1997). A nonparametric approach to estimating diver-gence times in the absence of rate constancy.Molecular Biology andEvolution, 14, 1218–1231.

Schenk, J. J., & Hufford, L. (2010). Effects of substitution models ondivergence time estimates: simulations and an empirical study ofmodel uncertainty using Cornales. Systematic Botany, 35, 578–592.

Sérsic, A., Cosacov, A., Cocucci, A. A., Johnson, L. A., Pozner, R., Avila,L. J., Sites Jr., J. W., & Morando, M. (2011). Emergingphylogeographical patterns of plants and terrestial vertebrates fromPatagonia. Biological Journal of the Linnean Society, 103, 475–494.

Smith, S. D., & Baum, D. A. (2006). Phylogenetics of the florally diverseAndean clade Iochrominae (Solanaceae). American Journal ofBotany, 93, 1140–1153.

Smith, S. D., Ané, C., & Baum, D. A. (2008). The role of pollinator shiftsin the floral diversification of Iochroma (Solanaceae). Evolution, 62,793–806.

Solà, E., Sluys, R., Gritzalis, K., & Riutort, M. (2013). Fluvial basinhistory in the northeastern Mediterranean region underlies dispersaland speciation patterns in the genus Dugesia (Platyhelminthes,Tricladida, Dugesiidae). Molecular Phylogenetics and Evolution,66, 887–888.

Starck, D., & Anzótegui, L. M. (2001). The late Miocene climaticchange—persistence of a climatic signal through the orogenic strat-igraphic record in northwestern Argentina. Journal of SouthAmerican Earth Sciences, 14, 763–774.

Stech, M., Veldman, S., & Larraín, J. (2013). Molecular species delimi-tation in the Racomitrium canescens Complex (Grimmiaceae) andimplications for DNA barcoding of species complexes in mosses.PloS One. doi:10.1371/journal.pone.0053134.

Strelin, M. M., Benitez-Vieyra, S., Ackermann, M., & Cocucci, A.(2016a). Flower reshaping in the transition to hummingbird pollina-tion in Loasaceae, subfam. Loasoideae despite absence of corollatubes or spurs. Evolutionary Ecology, 30, 401–407.

Strelin, M. M., Benitez-Vieyra, S., Fornoni, J., Klingenberg, C. P., &Cocucci, A. A. (2016b). Exploring the ontogenetic scaling hypoth-esis during the diversification of pollination syndromes inCaiophora (Loasaceae, subfam. Loasoideae). Annals of Botany,117, 937–947.

Swofford, D.L. (2003). PAUP*. Phylogenetic Analysis Using Parsimony(*and Other Methods). Version 4. Sinauer Associates, Sunderland,Massachusetts.

Tamura, K., Peterson, D., Peterson, N., Stecher, G., Nei, M., & Kumar, S.(2011). MEGA5: molecular evolutionary genetics analysis usingmaximum likelihood, evolutionary distance, and maximum parsi-mony methods. Molecular Biology and Evolution, 28, 2731–2739.

Särkinen, T., Pennington, R. T., Lavin, M., Simons, M. F., & Hughes, C.E. (2012). Evolutionary islands in the Andes: persistence and isola-tion explain high endemism in Andean dry tropical forests. Journalof Biogeography, 39, 884–900.

Weigend, M., & Gottschling, M. (2006). Evolution of funnel-revolverflowers and ornithophily in Nasa (Loasaceae). Plant Biology, 8,120–142.

Weigend, M., Gottschling, M., Hoot, S., & Ackermann, M. (2004). Apreliminary phylogeny of Loasaceae subfam. Loasoideae(Angiospermae: Cornales) based on trnL(UAA) sequence data, withconsequences for systematics and historical biogeography.Organisms Diversity & Evolution, 4, 73–90.

Weigend, M., Gröger, A., & Ackermann, M. (2005). The seeds ofLoasaceae subfam. Loasoideae (Cornales) II: seed morphology ofBSouth Andean Loasas^ (Loasa, Caiophora, Scyphanthus andBlumenbachia). Flora-Morphology, Distribution, FunctionalEcology of Plants, 200, 569–591.

Weigend, M., Ackermann, M., & Henning, T. (2010). Reloading therevolver- male fitness as a simple explanation for complex rewardpartitioning in Nasa macrothyrsa (Loasaceae, Cornales). BiologicalJournal of the Linnean Society, 100, 124–131.

Wen, J., & Zimmer, E. (1996). Phylogeny and biogeography of Panax L(the ginseng genus, Araliaceae): inferences from ITS sequences ofnuclear ribosomal RNA.Molecular Phylogenetics and Evolution, 6,167–177.

West-Eberhard, M. J. (2005). Developmental plasticity and the origin ofspecies differences. Proceedings of the National Academy ofSciences, 102, 6543–6549.

Xia, X., & Xie, Z. (2001). DAMBE: software package for data analysis inmolecular biology and evolution. Journal of Heredity, 92, 371–373.

Xia, X., Xie, Z., Salemi, M., Chen, L., & Wang, Y. (2003). An index ofsubstitution saturation and its application. Molecular Phylogeneticsand Evolution, 26, 1–7.

Yang, Z., & Rannala, B. (2012). Molecular phylogenetics: principles andpractice. Nature Reviews Genetics, 13, 303–314.

Zech, J., Zech, R., Kubik, P. W., & Veit, H. (2009). Glacier and climatereconstruction at Tres Lagunas, NWArgentina, based on 10 Be sur-face exposure dating and lake sediment analyses. Palaeogeography,Palaeoclimatology, Palaeoecology, 284, 180–190.

Zuloaga, F.O., Morrone, O., Belgrano, M.J., Marticorena, C., &Marchesi, E. (2008). Catálogo de las plantas vasculares del ConoSur. Monogr. Syst. Bot. Missouri Bot. Gard, 107(1–3), i–xcvi, 1–3348.

Diversification of Caiophora (Loasaceae subfam. Loasoideae) 41