Temporal origins and diversification of Artemisia and allies (Anthemideae, Asteraceae)

Temporal origins and diversification of Artemisia and allies (Anthemideae, Asteraceae).To assess temporal origins and diversification of lineages within subtribe Artemisiinae and Artemisia group a penalized likelihood analysis was applied on nrDNA ITS and eTS of 63 representatives. The tree was calibrated at the stem node of the Kaschgaria/ Artemisia lineage with the most reliable early Artemisia fossil pollen record from late Oligocene (23 Ma). The results from this study suggest that the origin of the subtribe goes back to the late Oligocene (24.6 ± 2.6 Ma) whilst the onset of differentiation of the genus Artemisia and most closely related genera is dated to the early Miocene (19.8 ± 2.3 Ma). Divergence ages for lineages within the Artemisia group are often between the early and Middle Miocene, whereas their radiations mostly occurred in the late Miocene and Pliocene. The temporal context was also used to examine biogeographic and morphological (capitula and pollen type) evolution. within the Artemisia group all lineages except the North American endemic have colonized the Mediterranean Basin at different epochs from Asian ancestors. Our analyses suggest the divergence of the North American endemic group from Asian ancestors (10.8 ± 1.5 Ma) in the late Miocene. homogamous-discoid capitula, characteristic of subgenera Seriphidium and Tridentatae, evolved not only in different geographic regions, but also at different times (2.0 ± 0.8 Ma and 7.9 ± 0.9 Ma respectively) within the Artemisia group. The loss of fertility of central flowers of disciform capitula should be considered as an ancient event in the genus since subgenus Dracunculus is one of the first groups that diverged (17.6 ± 2.1 Ma).

The first appearance of Artemisia has been widely discussed (Miao et al., 2011 and references therein).Pollen fossil records suggest that Artemisia originated in temperate Asia during the late eocene (Miao et al., 2011) or the Mid-Tertiary (wang, 2004).According to wang ( 2004), there are two reliable fossil pollen records of Artemisia in china, one from the early Oligocene in Xinjiang province (Song, 1965), the second from the late Oligocene in the Qinghai province (zhu et al., 1985;Miao et al., 2011).The age of the rocks for both places was originally determined by biostratigraphic evidence, but that of the Qinghai province was later adjusted by paleomagnetic data (cGTRP, 1992).The most ideal data are the fossil pollen sites with reliable paleomagnetic dating and high percentages of Artemisia pollen (Miao et al., 2011), therefore we considered the late Oligocene date (zhu et al., 1985) the most reliable early Artemisia fossil pollen record.Following ling (1991a, 1991b, 1994), the putative ancestor of Artemisia existed in Northern Asia and the migrations from this place supposedly occurred along three lines: (1) westward into europe, western Asia, Asia Minor, the Mediterranean Basin, and Africa; (2) eastward into Siberia and into western North America, and (3) further South into Asia.
The use of DNA sequences to estimate the timing of evolutionary events is increasingly popular.In plants, molecular dating has been used in numerous studies to investigate the timeframe of evolutionary events, e.g. for testing biogeographical hypotheses or to investigate the causes of recent radiations (reviewed in Sanderson et al., 2004).Dating techniques have been applied on taxa from very different taxonomic levels, e.g.angiosperms (Magallón & Sanderson, 2001;wikström et al., 2001;Bell et al., 2005), compositae (Funk et al., 2009), genera or species level (Park et al., 2006;Smith et al., 2008;Goodall-copestake et al., 2009).Oberprieler (2005)  Artemisia, Tkach et al. (2008b) dated the origins of Arctic lineages of the genus using a Bayesian approach with an uncorrelated lognormal clock on molecular eTS and ITS data.however, none of them have dealt with temporal evolution of lineages within the subtribe Artemisiinae and/or within the Artemisia group.
Therefore, the main goals of this study are (1) to assess temporal origins and diversification of the group, and (2) to examine biogeographic and morphological (capitula and pollen type) evolution of the group using this temporal context.

Data set
For the reconstruction of the temporal evolution of the group presented here we use the same data set as in our previous study (Sanz et al., 2008) that comprises sequence information for nrDNA ITS and eTS for 63 representatives of the subtribe.The data set consists of 45 species of Artemisia (46 populations) representing all five traditional subgenera and the geographical range of the genus, and 14 species from 13 genera belonging to subtribe Artemisiinae (Ajania Poljakov, Brachanthemum Dc., Dendranthema (Dc.) Des Moul., Elachanthemum Y. ling & Y.R. ling, Filifolium, Kaschgaria Poljakov, Mausolea, Neopallasia, Picrothamnus, Sphaeromeria, Turaniphytum, Nipponanthemum nipponicum Kitam.and Hippolytia megacephala (Rupr.)Poljakov.

Divergence time estimation
we used the most accurately dated and thus most reliable early Artemisia fossil pollen found in the late Oligocene in the Qinghai province (zhu et al., 1985) as calibration point.The age of this fossil cannot be assigned to any of the species within the Artemisia group and, hence, it can only be used to suggest a minimum age for Artemisia and closely related genera with Artemisia pollen type.considering the appearance of microechinate pollen in the lineage leading to Elachanthemum intricatum (Franch.)Y. ling & Y.R. ling as a convergence, the most suitable calibration point (Magallón, 2004) would be the stem node of Kaschgaria/Artemisia lineage.consequently, we fixed the age of the stem node of this clade at 23 million years ago (Ma), i.e. the Oligocene-Miocene boundary.
Previous maximum likelihood (Ml) analysis to infer tree topologies and branch lengths from the combined dataset was employed (Sanz et al., 2008).likelihood ratio test (lRT ;Felsenstein, 1988) for rate constancy indicated that neither ITS nor eTS evolved in a clock-like manner.hence, to obtain ultrametric trees we applied penalized likelihood (Pl; Sanderson, 2002) as implemented in r8s 1.7 (Sanderson, 2003) to estimate minimum ages of the most recent common ancestor (MRcA) of selected nodes.Pl is a semiparametric method that allows substitution rates to vary among lineages according to a smoothing parameter.The optimal smoothing parameter was chosen on the basis of the data by cross-validation (Sanderson, 2003).confidence intervals were calculated by data bootstrapped 100 times.we generated 100 phylograms topologically identical, but with different branch length and then was summarized age distribution for each particular node.The central 95% of the age distribution provides an interval of confidence (Sanderson & Doyle, 2001).Bootstrap data matrices were generated using the Seqboot program in Phylip package (Felsenstein, 1999).First, we obtained relative ages by arbitrarily setting the root age to an age of 100, which were subsequently translated into absolute ages by calibrating the stem node of the Kaschgaria/Artemisia lineage to 23 Ma based on fossil pollen record.All divergence time estimates are given in Ma and were placed into the current timescale for the cenozoic based on the International Union of Geological Sciences 2009 (IUGS) (Gibbard et al., 2009).The ultrametric tree was visualized in the program FigTree ver.1.1.2(Rambaut & Drummond, 2007).

RESULTS
likelihood ratio test for rate constancy rejected the assumption of a molecular clock for both data sets (p < 0.001).cross-validation analyses identified the optimal smoothing parameter for the combined data set as 2. Average divergence times plus standard deviations for selected nodes in the Ml combined tree are given in Table 1 and the chronogram obtained is shown in Fig. 1.

Temporal diversification
Following the calibration of the present tree (Fig. 1), the origin of the subtribe in Asia goes back to the late Oligocene (node A: 24.6 ± 2.6 Ma) whilst the onset of differentiation of the genus Artemisia and most closely related genera is dated to the early Miocene (node e: 19.8 ± 2.3 Ma).Divergence ages for the main lineages within the Artemisia group correspond mainly to the early or Middle Miocene, whereas their radiations mostly occurred in the late Miocene and Pliocene.Oberprieler (2005) reported the divergence time of subtribe Artemisiinae to be 7 -10 Ma (late Miocene) based on a molecular clock analyses of the tribe Anthemideae using ITS sequences.however, our dating estimates suggest that Artemisia fossil pollen was sparsely distributed during the early Miocene, and widely distributed during the late Miocene, being particularly abundant during the Pliocene and the Pleistocene (wang, 2004;Song et al., 2004;Miao et al., 2011).Furthermore, the present results are consistent with geologic and ecological changes during the late Tertiary in Asia (Axelrod et al., 1996), specifically with the onset of increased aridity in Asian climate by at least 22 Ma (as evidenced by paleomagnetic data) due to regional tectonic changes and the uplift of the Tibetan-himalayan complex (Guo et al., 2002).
The dry conditions in central Asia and the strong cooling of the late eocene, partly related to the above-mentioned tectonic events, probably favoured the Artemisia expansion (Miao et al., 2011).A similar pattern of explosive radiation in the Qinghai-Tibetan plateau within the last 20 Ma has been reported for other Asteraceae groups (Ligularia- Cremathodium-Parasenecio complex;liu et al., 2006).
According to our reconstructions, within the Artemisia group all lineages except the North American endemic group have colonized the Mediterranean Figure 1.chronogram of Artemisia and allied genera based on penalized likelihood analysis of a maximum likelihood tree constructed from combined nrDNA ITS and eTS data.Posterior probability values (above) and bootstrap supports (below) derive from our previous study (Sanz et al., 2008).To calibrate the tree, the stem node of the clade including Artemisia and Kaschgaria was fixed to 23 Ma.Nodes, for which ages were estimated, are indicated by capital letters.Geological epochs are abbreviated as follows: Plio, Pliocene; Pleis, Pleistocene.capitula types are mapped in different colors and taxa with Anthemis-type pollen are shown in grey.6.8 ± 0.8 Artemisia absinthium complex (h) 5.9 ± 1.5 Alpine group (I) 5.8 ± 0.8 Artemisia afra clade (J) 6.6 ± 1.6 Subgenus Seriphidium (K) 2.0 ± 0.8 North American group/A.vulgaris complex (l) 10.8 ± 1.5 North American group (M) 7.9 ± 0.9 Artemisia vulgaris complex (N) 7. Basin at different epochs from Asian ancestors.The first colonization of this area took place in the late Miocene during the Messinian salinity crisis (5.96-5.33Ma), when the closure of the Mediterranean-Atlantic gateways caused the nearly completely evaporation of the Tethys sea (hsü et al., 1973;Bocquet et al., 1978;Meulenkamp & Sissingh, 2003) and the progressive aridification of this region (Van Dam, 2006).examples include eastern and western Mediterranean species (A.judaica l., A. reptans c.Sm. ex link and A. lucentica O. Bolòs, Vallès-Xirau & Vigo) within the Artemisia/Absinthium clade (node G: 6.8 ± 0.8 Ma) and the ancestor leading to the A. absinthium complex (node h: 5.9 ± 1.5 Ma).later on, at the end of Pliocene there was a secondary colonization of the Mediterranean region by members of the subgenera Dracunculus (A. campestris l., A. monosperma Delile) and Seriphidium (A. barrelieri Besser, A. herba-alba Asso, A. inculta Delile).The dramatic changes experienced in the Mediterranean climate from 3.5 Ma onwards, with the stabilization of summer drought (Suc, 1984;Thompson, 2005) could facilitate the latter colonizations.
A westward colonization of the Mediterranean Basin is a common trend in plants (Paun et al., 2005;Mráz et al., 2007) and has been also described in other Anthemideae groups (lo Presti & Oberprieler, 2009).
Our dating results show that the divergence of A. canariensis less.from its continental sister species (node h: 5.9 ± 1.5 Ma), and the split of central Asian A. santolinifolia from the South African A. afra (node J: 6.6 ± 1.6 Ma) date back to late Miocene.These estimates are consistent with the increase of land bridges in the Mediterranean during the late Miocene (Meulenkamp & Sissingh, 2003) that also could have provided migration routes of the eurasian flora into Africa and Macaronesia as described in several other plant groups (Oberprieler, 2005;Mansion et al., 2008).
Our analyses suggest two arrivals to North America in the late Miocene.One concerns the divergence of the North American endemic group from Asian ancestors (node l: 10.8 ± 1.5 Ma) and the other the diversification of the American endemic A. mexicana willd.from its eurasian vicariants of the A. vulgaris l. complex (node N: 7.8 ± 1.3 Ma).Tkach et al. (2008b), based on uncorrelated lognormal clock method and calibrating the tree at the eocene/Oligocene boundary (34 Ma), obtained a similar divergence age of these groups (median age: 13.5 Ma and 7.5 Ma respectively).Floristic exchanges between Asia and North America were most likely across the Bering and the North Atlantic land bridges (Tiffney, 1985;wen, 1999).The Beringian land Bridge began to cool down from the Middle Miocene (15 Ma) onwards with an increasing cooling trend in the late Miocene (Mosbrugger et al., 2005;Milne, 2006) that would have favoured vicariance of Old and New world Artemisia species.In addition, our estimated age is consistent with a Middle Miocene (12 Ma) development of the sagebrush steppe in the Great Basin cold desert in North west America (Davis & ellis, 2010).
Our age estimates suggest an origin of the alpine group in the late Miocene (node I: 5.8 ± 0.8 Ma) and its diversification in the Pliocene.At that time, the old-tertiary Alpine orogeny had already formed the main european mountain belts (Smith et al., 1994).This finding supports the origin of orophytic groups within the genus earlier than Quaternary glaciations.Similar conclusions have been drawn for other european alpine taxa (comes & Kadereit, 2003;Paun et al., 2005).
Regarding temporal evolution of morphological traits, in the basal groups the divergence of the monotypic genus Nipponanthemum (node B: 12.1 ± 3.2 Ma) is a relatively ancient event in the subtribe compared to that of the Dendranthema species (node c: 4.3 ± 1.2 Ma).Therefore, the origin of radiate capitula and Anthemis pollen-type (both characters only found together in these two genera among the studied ones) occurred in different periods of time during the evolution of the subtribe.The reconstruction of the capitula type evolution within the Artemisia group (Sanz et al., 2008) suggests that all capitula types evolved from heterogamousdisciform capitula with central hermaphrodite and outer fe male, non radiate florets.Our age estimates indicate that Dracunculus (heterogamous-disciform capitula with central florets male) is one of the first groups that diverged (node F: 17.6 ± 2.1 Ma), therefore the loss of fertility of central florets of disciform capitula should be considered as an ancient event in the genus.In contrast, subgenus Seriphidium is the youngest group within Artemisia (node K: 2.0 ± 0.8 Ma) and it diverged independently of the North American subgenus Tridentatae that diverged approximately 6 million years earlier (node M: 7.9 ± 0.9).This implies that homogamous-Collectanea Botanica vol.30 (2011): 7-15, ISSN: 0010-0730, doi: 10.3989/collectbot.2011.v30.001discoid capitula (with all florets hermaphrodite), characteristic for both subgenera, evolved not only in different geographic regions, but also at different times within the Artemisia group.
Similarly, the evolution of the heterogamous capitula with male central florets, found in the lineage that gave rise to subgenus Dracunculus from perfect heterogamous capitula (Sanz et al., 2008) also arose in different periods in Asia and North America (Artemisia filifolia Torrey and the monotypic genus Picrothamnus).

CONCLUDING REMARKS
The results of this research suggest that the divergence ages for the main lineages within the Artemisia group chiefly correspond to the early and Middle Miocene, whereas the radiation of the groups occurred mostly in the late Miocene and Pliocene.
Our analyses suggest that all lineages, except the North American endemic group, have colonized the Mediterranean Basin at different epochs from Asian ancestors.The loss of fertility of central florets in Dracunculus from perfect heterogamous capitula is dated as an ancient event in the genus.
The scarce fossil data and the need of a broader sampling warrant caution in the interpretation of the molecular dating results, and further studies will be necessary to test our temporal hypotheses.

Table 1 .
Average divergence times (± standard deviation) in Ma of selected nodes (their designation by letters used in Fig.1is given in parentheses).