Phylogeny of Chaetanthera (Asteraceae: Mutisieae) reveals both ancient and recent origins of the high elevation lineages Mark A. Hershkovitz a,¤, Mary T.K. Arroyo a,b, Charles Bell c, L. Felipe Hinojosa a,b a Departamento de Ciencias Ecológicas, Facultad de Ciencias, Universidad de Chile, Casilla 653, Santiago, Chile Instituto de Ecología y Biodiversidad (IEB), Facultad de Ciencias, Universidad de Chile, Casilla 653, Santiago, Chile c School of Computational Science, 150-R Dirac Science Library, Florida State University, Tallahassee, FL 32306-4120, USA b Abstract Penalized likelihood analysis of previously published chloroplast DNA (cpDNA) ndhF sequences suggests that the central-southern Andean genus Chaetanthera diverged ca. 16.5 million years (my) ago, well before the uplift of the Andes to their present heights. Penalized likelihood analysis based on new nuclear ribosomal DNA (rDNA) internal transcribed spacer (ITS) sequences indicates that the most relictual lineages occupy high elevation Andean habitats that did not exist until some 10 my later. This result is contrary to the expectation that younger habitats should be occupied by phylogenetically younger lineages. The results are interpreted with respect to the development of aridity in lowland habitats during the Miocene and Pliocene, which presumably extinguished the lowland relatives of the high elevation taxa or, in eVect, forced them upwards in search of adequate moisture. As the more northerly lineages were being displaced upward, others diversiWed in the mediterranean-type climate area of central Chile, giving rise to additional high elevation taxa again, at an early date, as well as lowland taxa. Some species of Chaetanthera from lowland central Chile appear as the phylogenetically youngest taxa, suggesting secondary adaptation to lowland aridity. At the same time, at least two high elevation species, Chaetanthera peruviana and Chaetanthera perpusilla, appear to have been derived recently from a lower elevation ancestor, while some middle to low elevation taxa seem to have evolved recently out of a high elevation complex. The results suggest that the younger high elevation habitats have served as both “cradle” and “museum” for Chaetanthera lineages. Keywords: Chaetanthera; nrITS; Calibration age; Molecular clock 1. Introduction The central and southern Andes comprise a continuous high mountain range spanning from central Peru and Bolivia south through Argentina and Chile to Tierra del Fuego. This region harbors a diverse high elevation Xora of perhaps 3700 species (Arroyo, unpublished data) containing elements that are characteristic of, if not endemic to, South America, e.g. Calceolaria (Calceolariaceae; Erhart, 2000), Tropaeolum sect. Chilensia (Tropaeolaceae; Hershkovitz et al., 2006), Alstroemeria (Alstroemeriaceae; Muñoz Schick and Moriera Muñoz, 2003), and many gen* Corresponding author. Fax: +56 2 271 2983. E-mail address: mhershko@uchile.cl (M.A. Hershkovitz). era of Asteraceae tribe Mutisieae (Bremer, 1994). The high elevation Xora of the western slope of the Andes in Chile alone comprises an estimated 1700 species (Arroyo et al., 2004). For the present purposes, high elevation refers to the equivalent of above-treeline or alpine vegetation, a term that is diYcult to apply in central and northern southern Andes because of lack of a well-deWned treeline at the driest latitudes. In the northern part of the central Andes high elevation vegetation occurs between 3000 and 5000 m, descending to around 1800 and 3000 m heights at mediterranean latitudes in central Chile. Characterization as midelevation refers to plants below treeline or its equivalent along the slopes of the Andes, while lowland refers to coastal or low valleys below about 300–700 m elevation depending on latitude. M.A. Hershkovitz et al. Various lines of geological evidence indicate that uplifting of the central and southern Andes to their current elevations occurred between the Pliocene and Pleistocene (Arroyo et al., 1988; Farías et al., 2005; Gregory-Wodzicki, 2000; Giambiagi, 2003; Hartley, 2003; Irigoyen et al., 2000; Hinojosa and Villagrán, 1997; Simpson, 1983). Collectively, the data indicate that this region achieved far less to no more than one half of its current height by the end of the Miocene (Giambiagi, 2003; Gregory-Wodzicki, 2000). Assuming a correlation between the age of a habitat and the age of its biota, the geological data implicate a correspondingly recent origin of the central and southern high Andean Xora. In terms of phylogenetic trees, the expectation is that the stems of high elevation lineages should be relatively shorter than and nested among those of lower elevation lineages. Indeed, upward migration of lowland elements (Arroyo et al., 1983), along with recent long distance dispersal of high elevation plants established elsewhere (e.g. Soltis et al., 2001), have been proposed as important sources of the high Andean Xora. Among the taxa endemic to southern South America and well represented in both higher and lower elevations is Chaetanthera Ruiz & Pav. (Asteraceae: Mutisieae). As currently conceived, Chaetanthera comprises 44 species in seven subgenera (Cabrera, 1937). Subgenus Egania (perennial herbs) is restricted to high elevation habitats ranging from central Chile and Argentina to Peru and Bolivia. Subgenus Oriastrum (annuals to short-lived perennials) comprises high and high to mid-elevation species of central Chile and adjacent Argentina. Subgenus Carmelita (perennial herbs) comprise mostly high elevation species of central Chile and Argentina with occasional populations of some species occurring at mid-elevations. Subgenus Glandulosa (one subshrub) occurs in central Chile at mid-elevation to treeline. Subgenera Tylloma (mostly annual) and Euchaetanthera (all annuals) may be found from sea level to high elevations in central Chile, in the Andes in adjacent Argentina, and less frequently in northern Chile, Peru, Bolivia, and southern Argentina. Subgenus Proselia (perennials) is mostly concentrated in low to mid-elevation habitats, including in Araucaria forest clearings in south-central Chile, just getting into Argentina. Here, as part of a more comprehensive study into the evolution of Chaetanthera, we undertake phylogenetic reconstruction and estimation of divergence dates in order to provide insights into the evolution of the autochthonous central-southern Andean Xora in relation to the Andean uplift. We Wrst calibrate the age of Chaetanthera by comparison to other Asteridae for which divergence dates have been estimated. We then develop an ITS phylogeny for Chaetanthera and use the preceding calibration to estimate the divergence dates of the individual subgenera and relate these to the ecological history of the central–southern Andes. 2. Materials and methods Specimens of all but a few taxa of Chaetanthera were collected and portions preserved in silica gel for DNA analysis. Specimens were identiWed according to Cabrera (1937) and more recent publications on particular taxa. In addition, most identiWcations were veriWed by A. Davies (Munich, Germany), who is undertaking a revision of the genus in collaboration with the second author. In order to estimate the divergence date of Chaetanthera (see below), chloroplast DNA (cpDNA) ndhF sequences from Asteridae and Asteraceae listed in Kim et al. (2005) and partial ndhF sequences (3⬘ end, ca. 1000 bp) listed in Kim et al. (2002) were obtained from GenBank and aligned manually. In addition, the sequence of Cornus Xorida (GenBank Accession No. AF130220) was obtained. The alignment was trimmed to include only the 3⬘ end used in the Kim et al. (2002) analysis. Cornus was selected as a calibration point based on an approximate fossil dating, as in Bremer et al. (2004). Nuclear ribosomal DNA (rDNA) internal transcribed spacer (ITS) region sequences were obtained from 82 samples of Chaetanthera following the extraction, ampliWcation, and sequencing protocols described in Hershkovitz (2006). Eliminating 19 of the taxonomically duplicated samples, the present analysis is based on 63 samples (Table 1). Some samples did not amplify following the extraction protocol and were further puriWed using Chelex 100 resin (BioRad). 10 l of genomic DNA is added to 250 l of 5% aqueous Chelex, mixed, heated at 100 °C for 15 min, and the supernatant containing puriWed DNA removed to a new tube. This procedure removes heavy metal cations that apparently inhibited the PCR. Other puriWcation methods (PEG precipitation and/or silica-NaI) were tried, but they did not alleviate this inhibition. The resulting ITS sequences were aligned manually. The sequences included 56 apparently polymorphic sites, usually C/T, among the total 43,533 aligned sites. These sites were scored as ambiguities (“N”) in the phylogenetic analysis. The polymorphisms are mainly restricted to taxonomically diYcult species. Cloning of polymorphic samples is underway and the taxonomic and evolutionary implications will be considered in a future publication. Maximum parsimony (MP) and maximum likelihood (ML) analysis of the ndhF and ITS sequences were undertaken using PAUP 4.0 (SwoVord, 2002) version b10. The MP analysis of the ITS sequences included alignable gaps scored as separate characters. The MP analysis and bootstrap (500 replicates) were performed using the default heuristic search procedure. For the bootstrap analysis, maxtrees was set at 100. For the ML analysis, Modeltest (Posada and Crandall, 1998) was used to estimate the ML parameters. Tree rooting was based on Bremer et al. (2004) for Asteridae and on analysis of cpDNA rpl32-trnL intergenic spacer sequences for representative Chaetanthera species using Mutisia as the outgroup (data not shown). A likelihood ratio test was used to test for rate constancy among lineages (Felsenstein, 1981). In all cases, the hypothesis of rate constancy among lineages was rejected. Divergence dates were estimated over the ML topology and branch lengths using penalized likelihood (PL) implemented in the r8s program (Sanderson, 2002a,b). This M.A. Hershkovitz et al. Table 1 Taxa, vouchers, geographic origins, altitudes, and GenBank Accession numbers of specimens sampled for ITS. Vouchers are deposited in the Herbarium of the Universidad de Concepción (CONC) Subgenus species Carmelita C. lanata (Phil.) I. M. Johnst. C. spathulifolia Cabrera C. villosa D. Don C. villosa Egania C. acerosa (J. Rémy) Benth. & Hook. F. var. acerosa C. acerosa var. dasycarpa Cabrera C. acerosa var. indet. C. apiculata (J. Rémy) F. Meigen C. aV. boliviensis J. Kost. C. aV. cochlearifolia (Gray) B. L. Robinson C. dioica (J. Rémy) B. L. Robinson C. pentacaenoides (Phil.) Hauman C. pentacaenoides C. pulvinata (Phil.) Hauman var. pulvinata C. pulvinata var. pulvinata C. revoluta (Phil.) Cabrera C. sphaeroidalis (Reiche) Hicken C. aV. sphaeroidalis C. steubelii Hieron. var. abbreviata Cabrera C. steubelii var. argentina Cabrera C. steubelii var. indet. C. sp. ‘25203’ C. sp. ‘25204’ Euchaetanthera C. australis Cabrera C. chiquianensis Ferreyra C. ciliata Ruiz & Pav. C. euphrasioides (DC.) F. Meigen C. euphrasioides C. Xabellata D. Don C. incana Poepp. ex Less. C. leptocephala Cabrera C. linearis Poepp. ex Less var. albiXora Phil. C. linearis var. linearis C. linearis var taltalensis I. M. Johnst. C. linearis var taltalensis C. microphylla (Cass.) Hook. & Arn. C. moenchioides Less. C. perpusilla (Wedd.) Anderb. & S. E. Freire C. peruviana Gray C. tenella Less. var. taltalensis Cabrera C. tenella var. taltalensis C. tenella var. tenella Glandulosa C. glandulosa J. Rémy Oriastrum C. aV. gnaphalioides (J. Rémy) I. M. Johnst. C. aV. gnaphalioides C. lycopodioides (J. Rémy) Cabrera C. minuta (Phil.) Cabrera C. minuta C. planiseta Cabrera C. pusilla (D. Don) Hook. and Arn. Proselia C. brachylepis Phil. C. chilensis (Willd.) DC. var. chilensis C. chilensis var. tenuifolia (D. Don) Cabrera C. elegans (Phil.) var. elegans C. elegans var. pratensis (Phil.) Cabrera in M. N. Correa C. serrata Ruiz & Pav. Voucher Geographic origin Altitude (m) GenBank Accession Arroyo et al. 25075 Arroyo et al. 25098 Arroyo et al. 210671 Arroyo et al. 20646 CHILE: IV ARGENTINA: Mendoza CHILE: IX CHILE: RM 2790 3200 1300 2870 DQ355863 DQ355864 DQ355865 DQ355845 Arroyo et al. 25087A Arroyo et al. 25087B Arroyo et al. 25077 Arroyo et al. 25244 Arroyo et al. 25200 Arroyo et al. 25111 Arroyo et al. 25102 Arroyo et al. 25099 Arroyo et al. 25168 Arroyo et al. 25083 Arroyo et al. 25100 Arroyo et al. 25126 Arroyo et al. 25082 Arroyo et al. 25104 Arroyo et al. 25110 Arroyo et al. 25109 Arroyo et al. 25201 Arroyo et al. 25203 Arroyo et al. 25204 CHILE: IV CHILE: IV CHILE: IV CHILE: V BOLIVIA: La Paz ARGENTINA: Tucumán ARGENTINA: La Rioja ARGENTINA: Mendoza CHILE: RM CHILE: IV ARGENTINA: Mendoza CHILE: II CHILE: IV ARGENTINA: La Rioja ARGENTINA: Tucumán ARGENTINA: Tucumán BOLIVIA: Potosi CHILE: I CHILE: I 3700 3700 3210 2700 4811 4220 3340 3180 3310 4200 3030 4450 4360 5100 4220 3890 4300 4300 4850 DQ355909 DQ355914 DQ355905 DQ355910 DQ355911 DQ355895 DQ355898 DQ355893 DQ355904 DQ355903 DQ355909 DQ355899 DQ355897 DQ355913 DQ355900 DQ355896 DQ355912 DQ355901 DQ355913 Arroyo et al. 25177 Arroyo 25252 Arroyo et al. 25157 Arroyo et al. 25176 Arroyo et al. 25119 Arroyo et al. 25161 Arroyo et al. 25013 Hershkovitz 02–109 Arroyo et al. 25012 Arroyo et al. 25033 Arroyo et al. 25129 Arroyo et al. 25019 Arroyo et al. 25007 Arroyo et al. 25122 Arroyo et al. 25202 Arroyo 25254 Arroyo et al. 25055 Arroyo et al. 25128 Arroyo et al. 25006 ARGENTINA: Río Negro PERU: Ancash CHILE: VIII CHILE: RM CHILE: RM CHILE: RM CHILE: IV CHILE: III CHILE: IV CHILE: V CHILE: IV CHILE: IV CHILE: RM CHILE: IX CHILE: I PERU: Tacna CHILE: IV CHILE: II CHILE: RM 860 3500 100 3315 2470 2200 960 2152 950 760 ca. 50 1050 960 32 3370 3680 92 175 910 DQ355883 DQ355850 DQ355888 DQ355868 DQ355866 DQ355867 DQ355885 DQ355873 DQ355909 DQ355869 DQ355870 DQ355909 DQ355871 DQ355847 DQ355880 DQ355850 DQ355878 DQ355879 DQ355882 Arroyo et al. 25181 CHILE: RM 2430 DQ355881 Arroyo et al. 25086 Hershkovitz 02–154 Arroyo et al. 25169 Arroyo et al. 25079 Arroyo et al. 25127 Arroyo et al. 25120 Arroyo et al. 25180 CHILE: IV CHILE: IV CHILE: RM CHILE: IV CHILE: II CHILE: RM CHILE: RM 3500 1056 3140 3200 3500 2460 3360 DQ355908 DQ355906 DQ355920 DQ355890 DQ355907 DQ355892 DQ355916 Arroyo 25250 Arroyo 25229 Arroyo et al. 25042 Arroyo et al. 25069 Arroyo and Humaña 26000 Arroyo et al. 25131 CHILE: IX CHILE: VII CHILE: VII CHILE: VIII CHILE: IX CHILE: VIII 670 1800 1250 1470 1660 60 DQ355848 DQ355840 DQ355841 DQ355889 DQ355839 DQ355886 M.A. Hershkovitz et al. Table 1 (continued) Subgenus species Voucher Geographic origin Altitude (m) GenBank Accession Tylloma C. Xabellifolia Cabrera C. glabrata (DC.) F. Meigen C. glabrata (DC.) F. Meigen C. glabrata (DC.) F. Meigen C. kalinae Davies C. limbata (D. Don) Less. C. renifolia (J. Rémy) Cabrera C. splendens (J. Rémy) B. L. Robinson C. sp. ‘02–96’ Arroyo et al. 25078 Arroyo et al. 25130 Arroyo et al. 25163 Arroyo et al. 25065 Arroyo et al. 25076 Arroyo et al. 25150 Arroyo et al. 25175 Arroyo et al. 25084 Hershkovitz 02–96 CHILE: IV CHILE: II CHILE: RM CHILE: IV CHILE: IV CHILE: III CHILE: RM CHILE: IV CHILE: III 3200 145 2190 110 2740 828 3410 3250 2264 DQ355852 DQ355858 DQ355856 DQ355854 DQ355861 DQ355843 DQ355860 DQ355853 DQ355857 For geographic origin, roman numerals denote regions of Chile. RM refers to the Metropolitan Region. method has been shown to perform relatively well under simulation (Sanderson, 2002a; Ho et al., 2005) and has been widely used to estimate divergence time in plant lineages, including divergence dates of angiosperms as a whole. (e.g. Schneider et al., 2004; Bell et al., 2005). The optimal smoothing level was chosen via the cross validation procedure described by Sanderson (2002a). Divergences were calibrated using a date of 128 my for the crown Asteridae (Bremer et al., 2004). This, in turn, generated a divergence date estimate for the crown of the four included Chaetanthera samples. Finally, the estimated divergence date of the Chaetanthera crown based on the ndhF data was used to calibrate the divergence dates of the ITS sequences. 3. Results The MP bootstrap results of the Asteridae ndhF data are shown in Fig. 1. The consensus is poorly resolved relative to the Asteraceae phylogeny of Bremer et al. (2004) and with respect to Asteraceae tribal classiWcation in general (cf. Bremer, 1994). Consistent with earlier results, the tribe Barnedesioideae diverges Wrst from Asteraceae. The results for Chaetanthera are consistent with, but less resolved than those of Kim et al. (2002), who did not perform a bootstrap analysis. In particular, Chaetanthera is sister to Duidaea, although with only modest bootstrap support and keeping in mind that only 16 of the 76 genera of Mutisieae (Bremer, 1994) were sampled. Duidaea includes four shrubby species in Venezuela (Bremer, 1994). The hierarchical likelihood ratio Test (hLRT) procedure in Modeltest selected the TVM+G model, whereas the Aikake Information Criterion (AIC) selected the TVM+G+I model with corresponding compensation in the gamma parameter to account for the invariant site proportion estimate. A model more general than that derived from hLRT, the GTR+G model with estimated base frequencies, was used for ML analysis. Initial parameters were those estimated from a neighborjoining tree. Starting trees for ML analysis were those having the best ML score (18 trees) among the 3563 MP trees (length, L D 1704; rescaled consistency index, RC D 0.29; retention index, RI D 0.53). The Wnal ML topology (Fig. 2) was among the MP trees. The ultrametric tree derived from PL analysis, along with approximate divergence dates, are shown in Fig. 5. The MP bootstrap results of the ITS data for Chaetanthera are shown in Fig. 3. The data strongly support a major division of taxa at the base of the tree, one comprising subgenera Egania and Oriastrum (Clade A) and one comprising the remaining taxa (Clade B). We reiterate that rooting was based on cpDNA rpl32-trnL intergenic spacer sequences for representative Chaetanthera species using Mutisia as the outgroup (Hershkovitz and Arroyo, unpublished data). ITS sequences of three Mutisieae (Mutisia, Pachylaena, and Trichocline) could not be conWdently aligned with those of Chaetanthera, thus preventing outgroup rooting of the ITS data. However, midpoint rooting placed the root between the Egania and Oriastrum clades, reXecting the especially short branches of the former and long branches of the latter. This rooting would not impact the biogeographic or chronological interpretations. Both the hLRT and AIC procedures selected the TNM+G+I model. A more general GTR+G+I model with estimated base frequencies was used for ML analysis. Initial parameters were those estimated from a neighbor-joining tree. Starting trees for ML analysis were those having the best ML score (1 trees) among the 180 MP trees (L D 869, RC D 0.44, RI D 0.83). The ML analysis produced two similar topologies, the Wrst of which is shown in Fig. 4. This topology had an MP score of 872. The ML topology suggests that some species, as traditionally circumscribed, are para- or polyphyletic, including C. elegans, C. glabrata, C. tenella, C. minuta, C. steubelii, C. pulvinata, and possibly C. gnaphalioides and C. sphaeroidalis. The divergence between the two collections of C. villosa, one from central and the other from southern Chile, is also notable. Additional ITS data (not shown) suggest that C. moenchioides is also unnatural. The Wner details of Chaetanthera systematics will be the subject of a future paper. The ML tree (Fig. 4) suggests that evolutionary rates of ITS have been heterogeneous in Chaetanthera. The LRT yielded p D 0.002 against the hypothesis of rate constancy (i.e. the clock). The strongest rate contrast is between Egania and Oriastrum. This is possibly related to life form, as species of Egania are long-lived perennials, whereas species of Oriastrum are annuals or short-lived perennials. The ITS divergence rate of C. glandulosa, a subshrub, appears slower than in related taxa. M.A. Hershkovitz et al. Fig. 1. Maximum parsimony bootstrap consensus of Asteridae partial ndhF sequences. Bootstrap proportions are indicated adjacent to the branches. The divergence dating using PL (Fig. 5) suggests an origin of crown Asteraceae at ca. 36 my and an origin of crown Chaetanthera at ca. 16.5 my. This date separates subgenera Clades A and B and is in agreement with recent palynological evidence (Tellería and Katinas, 2004). Egania and Oriastrum subsequently diverge at ca. 13.5 my. The crown age of Clade B is ca. 10 my. The remaining subgenera, with the trivial exception of the monotypic subgenus Glandulosa, are nonmonophyletic, hence cannot be ascribed ages. With some exceptions, the higher elevation lineages appear to branch oV relatively early, with the lowest elevation taxa diverging most recently. The most recently derived high elevation taxon appears to be C. peruviana, which diverged from C. chiquianensis, a mid-elevation species, within the past million years and is nested within a mid- to low elevation clade. 4. Discussion The divergence dates among Asterales families based on partial ndhF sequences are in remarkably good agreement with those of Bremer et al. (2004) based on six chloroplast M.A. Hershkovitz et al. Fig. 2. Penalized likelihood estimatence of divergence dates of Asteridae based on partial ndhF sequences. sequences. The resulting estimate of ca. 16.5 my for the divergence of the crown group of Chaetanthera thus seems reasonably robust. The estimates for Asterales are somewhat diVerent from those of Kim et al. (2005) using diVerent methods. In particular, the latter shows Asterideae as much younger (according to their Fig. 6, apparently ca. 85 my), while Asteraceae are somewhat older (ca. 40 my) than in Bremer et al. (2004) and the present analysis. Thus, the discrepancy cannot be explained as a matter of scaling. It should be emphasized, however, that the Asteraceae datings generated by Kim et al. (2005) should yield an age estimate for Chaetanthera at least as old as that predicted by the Bremer et al. (2004) scaVold. Our age estimate for Chaetanthera is further supported by the ITS divergences. In Hershkovitz and Zimmer (2000), we estimated (based on Baum et al., 1998) a rate of 5 £ 10¡9 M.A. Hershkovitz et al. Fig. 3. Maximum parsimony bootstrap consensus of Chaetanthera ITS sequences. Bootstrap proportions are indicated adjacent to the branches. Numbers next to taxon names correspond to collection numbers, cf. Table 1. The traditional subgeneric classiWcation (Cabrera, 1937) of the taxa is shown at right. substitutions per site per year (ssy, Kimura 2-parameter, or K2P distances, unfortunately ignoring gamma for comparison to other published rates) for herbaceous plants. The Chaetanthera divergences would be at least slightly faster than this given our dates, which again points to the conservatism in our estimate for the age of the genus. For example, the C. Xabellata–C. elegans var. elegans split at ca. 10.5 my yields a rate of 7.6 £ 10¡9 ssy. The C. minuta–C. lycopodioides split at ca. 14 my yields a rate of 1 £ 10¡8 ssy, thus double the typical rate. The ML topology based on partial ndhF sequences is in remarkably good agreement with the tree based on combined complete ndhF and rbcL sequences (Kim et al., 2005). A notable exception is with respect to relations among the tribe Barnedesioideae, in particular Chuquiraga. This genus is sister to Dasyphyllum in the present analysis but sister to Doniophyon in Kim et al. (2005). This possibly represents a sequence submission error. A BLAST search (Altschul et al., 1990) indicated that the complete Chuquiraga (GenBank No. L39393) and Dasyphyllum (L39392) ndhF sequences are identical, whereas the rbcL sequence of the former (AY874427) is more similar to that of Doniophytum (AY874430). Yet, ndhF sequences evolve more rapidly than rbcL. We leave this M.A. Hershkovitz et al. Fig. 4. Maximum likelihood phylogram of Chaetanthera ITS sequences. One of two topologies of equal likelihood. Numbers next to taxon names correspond to collection numbers, cf. Table 1. problem unresolved, however, because it does not aVect our general conclusions. There appears to be some discrepancy between the ndhF (Figs. 1 and 2) and ITS data (Figs. 3–5) concerning the relations among Chaetanthera species. In particular, the former shows strong bootstrap support for a clade comprising C. pusilla and C. Xabellifolia relative to C. acerosa. The ITS bootstrap (Fig. 3) and our cpDNA rpl32-trnL intergenic spacer sequences (data not shown) support the latter two as closer relatives. This makes better sense considering the respective subgeneric placements of the three species. Chaetanthera pusilla and C. acerosa belongs to Clade A without any doubt, whereas C. Xabellifolia belongs to Clade B. Assuming that the specimens are correctly determined in both cases, the source of the discrepancy is not clear. It is possible that poor sampling for ndhF yields a branch attraction artifact, or it is possible that the ITS tree is misrooted. If indeed the relations shown in the ndhF tree are M.A. Hershkovitz et al. Fig. 5. Penalized likelihood estimate of divergence dates of Chaetanthera based on ITS sequences. Numbers next to taxon names correspond to collection numbers, cf. Table 1. Altitudinal ranges are indicated as follows: white dots, low elevation; hatched dots, mid-elevation; black dots, high elevation. correct, the divergence date of Egania would increase to 16.5 my. The ultrametric tree with estimated divergence dates indicate that the high elevation subgenus Egania, centered in the central Andes and northern part of the southern Andes, diverged from the remainder of Chaetanthera at least 13.5 my ago, at which time the Andes had uplifted to perhaps half of their current height (Gregory-Wodzicki, 2000). However, Fig. 5 illustrates that the diversiWcation of modern taxa of Egania occurred during the past 5 my, in agreement with the age of the highest modern surfaces of the Andes. This pattern might have been produced by high extinction rates between 13.5 and 5 my ago of lower-altitude adapted plants in this lineage. Alternatively, this lineage may have been represented by a single species prior to a Pliocene radiation. This situation contrasts with that of subgenus Oriastrum, for which the lineages giving rise to the modern high-elevation taxa diversiWed prior to the Pliocene. M.A. Hershkovitz et al. The apparently early divergence of the lineages giving rise to Egania and Oriastrum begs the question as to what habitat the ancestor could have occupied, considering that the modern habitats of these taxa did not exist at that time. For that matter, what was the habitat of the ancestor of Chaetanthera and what circumstances account for the initial split between the Egania plus Oriastrum and the remainder of the genus? Existing geological evidence seems to rule out the possibility that the ancestors referred to above were already adapted to the high elevation environment. Although the sampling for Mutisieae remains sparse, there is no indication that Chaetanthera is closely related to any of the other genera with high elevation species, e.g. Nassauvia, Perezia, or Leucheria. As a whole, and notwithstanding the high incidence in the Andes, the tribe is considered to be pantropical and low elevation (Bremer, 1994). That the genus did not migrate from high elevation habitats in the northern Andes is ruled out by the apparently younger age of the northern relative to the central and southern Andes (Gregory-Wodzicki, 2000). The distribution of Mutisieae also seems to rule out the possibility that Chaetanthera migrated from a high elevation zone outside of South America. A more plausible explanation for apparent relictuality of Egania and Oriastrum is the increasing aridity from coastal Peru through the Atacama region to north-central Chile during the Pliocene (Hartley and Chong, 2002). Existing evidence suggests that aridity emerged in the Atacama region as early as the Jurassic, making it perhaps the world’s oldest desert (Hartley et al., 2005). However, hyperaridity did not develop until the late Pliocene. Thus, the Atacama region could have supported the existence of the Chaetanthera lineages prior to the Pliocene. These ancestors apparently escaped hyperaridity by elevating their habitat as the Andean uplift gradually proceeded. According to Kœrner (1999), higher elevation habitats are less water stressed than lower elevation habitats, in part because of lower evapotranspiration resulting from a shorter growing season and colder temperatures. One consequence is that soil moisture, especially deeper in the proWle, remains higher than at low elevation locations with comparable precipitation. In the case of the central and southern Andes to around 38 °S, precipitation is always higher at the highest elevations. This is the case for moisture derived from both the westerlies in central and southern Chile and the “invierno boliviano” easterlies that characterize the Andes of northern Chile, Bolivia, and Peru (Schwerdtfeger, 1976). Moreover, precipitation at high elevations is largely in the form of snow, which allows release of available water for an extended period. The principal innovation required for the evolution of the high elevation Chaetanthera species would be not for drought, but for cold tolerance (Kœrner, 1999). A variety of morphological and physiological mechanisms could result in this tolerance, among these the formation clusters of subterranean perennating buds and extensive Wne roots (cf. Kœrner, 1999), as seen in all species of Egania. This lineage subsequently diversiWed to occupy the high elevation zones receiving both winter precipitation (central Chile) and summer precipitation (the altiplano). Mechanisms of cold tolerance in the annual species of Oriastrum are less clear. However, the arid Andes in Chile stand out for their large number of annual species derived from many diVerent genera, with some (e.g. Chaetanthera pusilla) reaching the upper limits of the vegetation. The long, sunny growing season in the arid Andes of central Chile does not exclude annuals from successfully completing their life cycle (Arroyo et al., 1981). In any case, physiological mechanisms for high elevation cold tolerance in Chaetanthera have not been examined. Of equal importance is the biogeographic explanation for the divergence pattern of Clade B. These taxa are more decidedly central Chilean compared to Egania, with the exception of two closely related annual species (C. chiquianensis, C. peruviana) that occur in Peru and one (C. perpusilla) that occurs in Bolivia and Chile. The ancestor of this group is dated at ca. 11 my, which is still older than the high elevation habitats that some of the species occupy today, e.g. C. Xabellata, C. euphrasioides, C. spathulifolia and C. villosa are high elevation species. Thus, it seems that a mechanism similar to that proposed for Egania and Oriastrum is again required to explain the age of the high elevation lineages, i.e. late Miocene aridity in central Chile driving some plants upwards. Geological evidence shows that between the latitudes 33°S and 35 °S, processes of Andes deformation began between 15 and 16 my, followed by uplift beginning at ca. 9 my (Irigoyen et al., 2000; Giambiagi, 2003; Farías et al., 2005). Uplift would have aVected the circulation of humid air masses, coming from the Atlantic and the PaciWc. This would have resulted in a major reduction in summer rain on the Chilean side of the Andes, along with a reduction in winter rain on the Argentinian side. Consequently, a subtropical xeric Xora developed on opposite sides of the Andes (Hinojosa and Villagrán, 1997; Hinojosa, 2005; Villagrán and Hinojosa, 2005), the precursor to the present mediterranean matorral on the western side of the Andes and the Monte on the eastern side. The change from C3 to C4 grass domination that occurred between 7 and 4 my (Latorre et al., 1997, cf. Hartley and Chong, 2002) would be indirect evidence for the establishment of arid conditions at the end of the Miocene and beginning of the Pliocene as a result of the rain shadow produced by the Andes. The only clear exceptions to the biogeographic scenario proposed above are the high elevation species C. peruviana, and C. perpusilla, the Wrst of which is closely related and probably recently diverged from C. chiquianensis, a lower elevation species, and is nested in a predominantly mid- to low elevation clade. Although Table 1 shows the elevation of the collection of C. peruviana as only slightly higher than that of C. chiquianensis, the latter was collected further north. As we have noted, the limit of southern hemisphere alpine vegetation increases with decreasing latitude. In any case, C. chiquianensis was collected in a columnar cactusshrubland vegetation, whereas C. peruviana was collected M.A. Hershkovitz et al. ca. 300 m above the lower limit of the alpine puna vegetation. Chaetanthera perpusilla is a high elevation species whose closest relative is the coastal desert species C. tenella var. taltalensis, which is found in disjunct populations along the coast, like many other arid mediterranean-type climate species. Because this clade is strongly central Chilean, it seems that long distance dispersal is the most likely explanation for the occurrence of these species in high elevation habitats. This seems all the more likely given the relatively easy dispersal of the hairy achenes of Chaetanthera and in view of apparently frequent longer-distance amphitropical dispersal of taxa with no evident dispersal mechanisms, e.g., Portulacaceae (Hershkovitz and Zimmer, 2000). In any case, C. peruviana and C. perpusilla, both small annuals, are notable in that they are some of the few species of Chaetanthera whose phylogenetic relations correspond to the conventional wisdom that the younger high elevation habitat is occupied by a species derived from a lower elevation older habitat. Another striking aspect of Fig. 5 is that the generally lowest elevation species of Chaetanthera of Clade B appear to be the most evolutionarily derived, contrary to intuition. The above historical biogeographic scenario may provide some insight as to why this is so. The lower elevation taxa would have required a secondarily evolved tolerance to the increased aridity developing on the western slope of the Andes that became intense from the Pliocene onwards, giving rise to open mediterranean-type climate shrublands, which at the same time provided new niches for the establishment of an herbaceous Xora. It is noteworthy in this regard that the low-mid elevation species classiWed by Cabrera (1937) in subgenus Proselia found in south-central Chile often occupy open sites in volcanic soils. The annual species classiWed by Cabrera in Tylloma and Euchaetanthera generally occur in washes, gravelly slopes and Xats, and in well-drained open spaces between woody matorral shrubs, all of which become exceedingly dry in summer. The degree to which the evolution of high-elevation Chaetanthera proves to be the exception or the rule among other Andean taxa remains to be determined, Apparent relictuality of high elevation lineages appears to be the case in Portulacaceae and Tropaeolum sect. Chilensia. In the former, divergence of western American Portulacaceae genera appears to have occurred rather abruptly (Hershkovitz, 2006). Among the lineages that diverged are the high elevation or likely ancestrally high elevation taxa Montiopsis, Lenzia, and Calandrinia. Based on minimal ITS divergence of 0.08% (K2P distance) between western American Portulacaceae and the outgroup Phemeranthus, and using a substitution rate of 5 £ 10¡9 substitutions per site per year, an age of 16 my also emerges. In Tropaeolum sect. Chilensia the high-elevation clade including T. polyphyllum diverges immediately following separation of the clade including T. speciosum and the Argentinian disjunct, T. pentaphyllum, a disjunction probably late Eocene to Oligocene in age (Hershkovitz et al., 2006). ITS sequence divergence of the highaltitude clade from the more derived lowland taxa is on the order of 0.06%, but species of Tropaeolum are more longerlived perennials than most taxa of Chaetanthera and Portulacaceae, hence might have a slower rate of ITS divergence. In addition to the cases above, Fig. 2 also shows relictuality of another high-elevation genus, Pachylaena, as well as early divergence of other genera that include high elevation species, e.g. Perezia and Leucheria. These results cannot be considered as conclusive regarding high-elevation plant divergence time as the results for Chaetanthera. In particular, the genera of Mutisieae remain poorly sampled at the molecular level (see Bremer, 1994). Likwise, Perezia and Leucheria include many low elevation species. Comparison of K.-J. Kim et al. (2005) and the present results for the high elevation genus Nassauvia reveals how poor sampling can be misleading. The former suggest a divergence of ca. 30 my for this genus, whereas, with the inclusion of the Triptilion sample, the divergence is relatively recent. However, the present sampling cannot discriminate between a recent divergence of the high elevation Nassauvia relative to the lowland Triptilion versus the reverse scenario. The above discussion presents a new angle on the infamous “cradle versus museum” controversy regarding angiosperm origins (Stebbins, 1974), viz. whether angiosperms originated in tropical lowland forests where relictual species are overrepresented versus whether they originated in harsher habitats and evolved subsequently into habitats more favorable for luxurient growth. In the present case, the relatively young high altitude habitat appears to be a “museum” for some early derived lineages of Chaetanthera. In this particular case, the emerging high elevation habitat was hydrologically more favorable for growth than the increasingly arid lower elevation habitats in which the genus must have originated. However, as Stebbins (1974, p. 14), noted, a given habitat is probably an evolutionary laboratory for some taxa and a museum for others. This appears to be the case, as C. peruviana and C. perpusilla appear to have become adapted the high elevation habitat more directly and recently. Acknowledgments Research supported by FONDECYT Grant 1020956, an A. W. Mellon Foundation Grant, Millennium ScientiWc Initiative (ICM) Grant P02-051, and BBVA Foundation Prize in Conservation Biology Research. We acknowledge the able technical assistance of Ana María Humaña, Christian Zavaleta, Sebastian Molinet, and Gabriel Pérez. We thank Alison Davies for assisting in the identiWcation of most of the taxa. References Altschul, S.F., Gish, W., Miller, W., Myers, E.W., Lipman, D.J., 1990. Basic local alignment search tool. J. Mol. Biol. 215, 403–410. Arroyo, M.T.K., Armesto, J.J., Villagrán, C., 1981. Plant phenological patterns in the high Andean Cordillera of Central Chile. J. Ecol. 69, 205–233. M.A. Hershkovitz et al. Arroyo, M.T.K., Armesto, J., Primack, R., 1983. Tendencias altitudinales y latitudinales en mecanismos de polinización en los Andes templados de Sudamérica. Rev. Chil. Hist. Nat. 56, 159–180. Arroyo, M.T.K., Squeo, F.A., Armesto, J.J., Villagrán, C., 1988. EVect of aridity on the plant diversity of the northern Chilean Andes: results of a natural experiment. Ann. Missouri Bot. Gard. 75, 55–78. Arroyo M.T.K., Squeo F., Cavieres L., Marticorena, C., 2004. Chilenische Anden. In: Burga, C.A., Kötzli, F., Grabherr, G. (Eds.), PXanzenwelt der Gerbirg. Landschaft, Klima, PXanzenwelt. Ulmer, Stuttgart, pp. 210–219. Baum, D.A., Small, R.L., Wendel, J.F., 1998. Biogeography and Xoral evolution of baobabs (Adansonia, Bombacaceae) as inferred from multiple data sets. Syst. Biol. 47, 181–207. Bell, C.D., Soltis, D.E., Soltis, P.S., 2005. The age of the angiosperms: a molecular timescale without a clock. Evolution 59, 1245–1258. Bremer, K., 1994. Asteraceae: Cladistics and ClassiWcation. Timber Press, Portland, Oregon. Bremer, K., Friis, E.M., Bremer, B., 2004. Molecular phylogenetic dating of asterid Xowering plants shows early Cretaceous diversiWcation. Syst. Biol. 53, 496–505. Cabrera, A., 1937. Revisión del género Chaetanthera (Compositae). Rev. Mus. La Plata, Sección Botánica 1, 87–215. Erhart, C., 2000. Die Gattung Calceolaria in Chile. Bibliotheca Botanica, 153. E. Schweizerbart’she, Stuttgart. Farías, M., Charrier, R., Fock, A., Campbell, D., Martinod, J., Comte, D., 2005. Rapid late Cenozoic surface uplift of the central Chile Andes (33°–35 °S). In: 6th International Symposium on Andean Geodynamics, Barcelona, Spain (Abstract), pp. 261–265. Felsenstein, J., 1981. Evolutionary trees from DNA sequences: a maximum likelihood approach. J. Mol. Evol. 17, 368–376. Giambiagi, L.B., 2003. Deformación cenozoica de la faja plegada y corrida del Aconcagua y Cordillera Frontal, entre los 33°30⬘ y 33°45⬘S. Rev. Asoc. Geol. Argentina 58, 85–96. Gregory-Wodzicki, K.M., 2000. Uplift history of the central and northern Andes: a review. Geol. Soc. Amer. Bull. 112, 1091–1105. Hartley, A.J., 2003. Andean uplift and climate change. J. Geol. Soc. London 160, 7–10. Hartley, A.J., Chong, G., 2002. Late Pliocene age for the Atacama Desert: implications for the desertiWcation of western South America. Geology 30, 43–46. Hartley, A.J., Chong, G., Houston, J., Mather, A.E., 2005. 150 million years of climatic stability: evidence from the Atacama Desert, northern Chile. J. Geol. Soc. London 162, 421–424. Hershkovitz, M.A., 2006. Ribosomal and chloroplast DNA evidence for diversiWcation of western American Portulacaceae in the Andean region. Gayana. Bot., in press. Hershkovitz, M.A., Hernández-Pellicer, C.C., Arroyo, M.T.K., 2006. Ribosomal DNA evidence for the diversiWcation of Tropaeolum sect. Chilensia (Tropaeolaceae). Plant Syst. Evol., Available from: <http:// dx.doi.org/10.1007/s00606-006-0428-7>. Hershkovitz, M.A., Zimmer, E.A., 2000. Ribosomal DNA evidence and disjunctions of western American Portulacaceae. Mol. Phylogen. Evol. 15, 419–439. Hinojosa, L.F., 2005. Climatic and vegetational changes inferred from Cenozoic Southern South America paleoXora. Rev. Geol. Chile 32, 95–115. Hinojosa, L.F., Villagrán, C., 1997. History of the southern South American forests .1. Paleobotanical, geological and climatical background on Tertiary of southern South America. Rev. Chil. Hist. Nat. 70, 225–239. Ho, S.Y.W., Phillips, M.J., Drummond, A.J., Cooper, A., 2005. Accuracy of rate estimation using relaxed-clock models with a critical focus on the early metazoan radiation. Mol. Biol. Evol. 22, 1355–1363. Irigoyen, M.V., Buchan, K.L., Brown, R.L., 2000. Magnetostratigraphy of Neogene Andean foreland-basin strata, Lat 33 °S, Mendoza Province, Argentina. Geol. Soc. Amer. Bull. 112, 803–816. Kim, H.-G., Loockerman, D.J., Jansen, R.K., 2002. Systematic implications of ndhF sequence variation in the Mutisieae (Asteraceae). Syst. Bot. 27, 597–609. Kim, K.-J., Choi, K.-S., Jansen, R.K., 2005. Two chloroplast DNA inversions originated simultaneously during the early evolution of the sunXower family (Asteraceae). Mol. Biol. Evol. 22, 1783–1792. Kœrner, C., 1999. Alpine Plant Life. Springer, Berlin. Latorre, C., Quade, J., Macintosh, W.C., 1997. The expansion of C4 grasses and global change in the late Miocene: stable isotope evidence from the Americas. Earth Planet. Sci. Lett. 146, 83–96. Muñoz Schick, M., Moriera Muñoz, A., 2003. Alstroemerias de Chile. Taller La Era, Santiago. Posada, D., Crandall, K.A., 1998. Modeltest: testing the model of DNA substitution. Bioinformatics 14, 817–818. Sanderson, M.J., 2002a. Estimating absolute rates of molecular evolution and divergence times: a penalized likelihood approach. Mol. Biol. Evol. 19, 101–109. Sanderson, M.J., 2002b. r8s 1.50. Computer program and documentation. Available from: <http://phylo.ucdavis.edu/r8s/r8s.html/>. Schneider, H., Schuettpelz, E., Pryer, K.M., CranWll, R., Magallón, S., Lupia, R., 2004. Ferns diversiWed in the shadow of angiosperms. Nature 428, 553–557. Schwerdtfeger, W., 1976. Climates of Central and South America. In: Schwerdtfeger, W. (Ed.), World Survey of Climatology. Elsevier, Amsterdam, p. 145. Simpson, B.B., 1983. An historical phytogeography of the high Andean Xora. Rev. Chil. Hist. Nat. 56, 109–122. Soltis, D., KuzoV, R.K., Mort, M.E., Zanis, M., Fishbein, M., HuVord, L., Koontz, J., Arroyo, M.T.K., 2001. Elucidating deep-level phylogenetic relationships in Saxifragaceae using sequences for six chloroplastic and nuclear DNA regions. Ann. Missouri Bot. Gard. 88, 669–693. Stebbins, G.L., 1974. Flowering Plants: Evolution above the Species Level. Belknap Press, Harvard University, Cambridge, Massachusetts. SwoVord, D.L., 2002. PAUP*: phylogenetic analysis using parsimony (* and other methods), version 4. Sinauer Associates, Sunderland, Massachusetts. Tellería, M.C., Katinas, L., 2004. A comparative palynological study of Chaetanthera (Asteraceae, Mutisieae) and allied genera. Syst. Bot. 29, 752–773. Villagrán, C., Hinojosa, L.F., 2005. Esquema biogeográWco de Chile. In: Llorente, J., Morrone, J.J. (Eds.), Regionalización BiogeográWca en Iberoamérica y Tópicos AWnes: Primeras Jornadas BiogeográWcas de la Red Iberoamericana de Biogeografía y Entomología Sistemática. Las Prensas de Ciencias, UNAM, Mexico City, pp. 551–557.