Appendix S1 Construction of plant phylogeny.
DNA extraction, PCR amplification and sequencing
The silica-gel dried leaf materials which were collected at the same time with the root
samples were used for DNA extraction of each species. DNA was extracted by plant
genomic DNA kit (TIANGEN, Beijing) according to the protocol. Then partial
chloroplast gene fragments rbcL and matK were amplified. Program for polymerase
chain reaction (PCR) was 94 ºC for 5 min; then 35 cycles of 94 ºC for 30 sec, 50 ºC for
45 sec, 72 ºC for 45 sec; followed by a final extension of 72 ºC for 8 min. Primers for
the target gene fragments (Li et al., 2011) were listed in Table S2. Unpurified PCR
products were sequenced from both directions by Sangon using an
ABI-PRISM3730XL Genetic Analyzer (Foster City, CA, USA) with
BigDyeterminator v3.1.
Sequences assembling and alignment
Sequences of both directions were assembled for each species using SeqMan in
Lasergene 7.1 (DNASTAR, http://www.dnastar.com/). Only samples which
two-direction chromatograms matched above 97% were used in next analysis. MEGA5
(Tamura et al. 2011) was used to align the sequences with default set. Then the results
of alignment were manually corrected.
Construction of phylogenetic tree
As phylogenetically related outgroups give robust topology (Rosenfeld et al. 2012),
seven species which are related to the 85 sampled species were selected as outgroups
according to ANITA grade and APGIII (APG 2009; Doyle 2012). The outgroup
sequences were downloaded from Genbank. They were Amborella trichopoda (L12628;
1
AF543721), Nymphaea odorata (M77034; AF092988), Brasenia schreberi (M77031;
HQ189138), Illicium floridanum (DQ182334; AF543738), Trimenia moorei
(AY116658; DQ401360), Austrobaileya scandens (L12632; DQ182344), and
Schisandra chinensis (AF238061; JF956214).
Phylogenetic trees based on the combined dataset were constructed by maximum
likelihood and Bayesian approach. jModelTest v2.1.1 (Posada 2008) was used to
decide the best substitution model for both approaches. The best model was TVM+I+G.
For each gene fragment the best model was also TVM+I+G.
The maximum likelihood trees were constructed by PhyML3.0 (Guindon &
Gascuel 2003). Substitution model, equilibrium frequencies, relative rate parameters
and Gamma distribution parameter were all referring to the result of jModelTest.
BioNJ tree was used as starting tree. The best nearest neighbor interchange (NNI) and
subtree pruning and regrafting (SPR) were used to search tree topology with the
heuristic search which adds 5 random starting trees. One hundred non-parameter
bootstraps were used.
Bayesian trees were constructed by BEAST1.7.1 (Drummond et al. 2012) at the
same time of divergence time estimation. Details are shown in divergence time
estimation parts.
Estimation of divergence time
Data were partitioned into two parts (1+2), 3 based on codon position. Under the best
substitution model (TVM+I+G), 5 scenarios of prior distribution of fossil calibrated
nodes (details see below), 4 tree prior (Birth-Death Process, Birth-Death incomplete
sampling, Yule Process, Calibrated Yule) and 3 clock models (strict clock, uncorrelated
exponential relaxed clock, uncorrelated lognormal relaxed clock) were compared. As
2
several recent research estimated that the origin of angiosperm were around 200 Mya
(Bell et al. 2010; Magallón et al. 2013), and the first credible fossil of angiosperm is
about 130 Mya, thus the root height were set as normal distribution which 95%
confidence interval from 130 to 200 Mya. Bayesian factor (BF) was used as a criterion
of model selection.
Eight fossil calibrated nodes referring to Magallón & Castillo (2009) were used.
Detailed fossil information is shown in Table S4. Ages of fossils referred to GTS2012
(Cohen 2012; Gradstein 2012). Five scenarios were set as prior distribution for these 8
calibrated nodes as follows:
1) All calibrated nodes were given exponential distribution which mean value
equal to 1 and offset equal to the lower limit of fossil age;
2) All calibrated nodes were given lognormal distribution with the offset equal to
the lower limit of fossil age and the logstdev equal to 0.5, logmean was set to
make medium at the upper limit of fossil age;
3) All calibrated nodes were given normal distribution with the 95% confidence
interval equal to the interval between upper and lower limits of fossil age;
4) Exponential distribution for calibrated nodes of most recent common ancestor
(MRCA) of Fagales and Eudicots as the fossils are pollen (Sauquet et al. 2012)
and lognormal distribution for the rest 6 nodes, setting in the same way in
scenario 1) and 2);
5) Exponential distribution for calibrated nodes of MRCA of Fagales and
Eudicots and normal distribution for the rest 6 nodes, setting in the same way
in scenario 1) and 3).
After 6 to 50 million generations for each independent Markov Chain Monte Carlo
(MCMC), ensuring effective sample size (ESS) of every parameter over 100, which
3
indicates MCMC converged, Bayes factor was calculated by Tracer v1.5
(http://tree.bio.ed.ac.uk/software/tracer/). As the 85 studied species are unequally
sampled in different orders, just like extinction happened. Thus we tested combination
of different calibrated scenarios and clock models when Birth-Death Process was set as
tree prior first. After the best combination of calibrated prior and clock model was
decided, other 3 prior trees were compared under the best model combination. Bayes
factor support Birth-Death Process as the best tree prior. Then another run was taken
for every combination of calibrated scenario and clock model under Birth-Death tree
prior. After finding the best model combination, combined the results of 2 independent
run by LogCombiner v1.7.1 excluding first 10% burn-in, then maximum clade
credibility tree with mean heights was annotated by Tree Annotator v1.7.1.
Results
Both maximum likelihood and Bayesian approaches gave almost the same topology,
except in two nodes which bootstrap or posterior probability is low. This topology is
similar to the angiosperm tree built by Magallón & Castillo (2009) and APGIII (APG
2009). The best model combination for divergence time estimating is TVM +I+G for
substitution model, lognormal distribution for all eight fossil calibrated nodes,
birth-death process for tree prior and uncorrelated exponential relaxed clock for clock
model.
4
REFERENCES
1.
APG. (2009). an update of the angiosperm phylogeny group classification for the orders
and families of flowering plants: APG III. Bot. J. Linn. Soc., 161, 105-121.
2.
Bell C.D., Soltis D.E. & Soltis P.S. (2010). The age and diversification of the
angiosperms re-revisited. Am. J. Bot., 97, 1296-1303.
3.
Doyle J.A. (2012). Molecular and fossil evidence on the origin of angiosperms. Annu.
Rev. Earth Pl. Sc., 40, 301-326.
4.
Drummond A.J., Suchard M.A., Xie D. & Rambaut A. (2012). Bayesian phylogenetics
with BEAUti and the BEAST 1.7. Mol. Biol. Evol., 29, 1969-1973.
5.
Guindon S. & Gascuel O. (2003). A simple, fast, and accurate algorithm to estimate
large phylogenies by paximum likelihood. Syst. Biol., 52, 696-704.
6.
Li, D.Z., Gao, L.M., Li, H.T., Wang, H., Ge, X.J., Liu, J.Q., Chen, Z.D., Zhou, S.L.,
Chen, S.L. & Yang, J.B. (2011). Comparative analysis of a large dataset
indicates that internal transcribed spacer (ITS) should be incorporated into the
core barcode for seed plants. Proc. Natl. Acad. Sci., 108, 19641-19646.
7.
Magallón S. & Castillo A. (2009). Angiosperm diversification through time. Am. J. Bot.,
96, 349-365.
8.
5
Magallón, S., Hilu, K.W. & Quandt, D. (2013). Land plant evolutionary timeline: gene
effects are secondary to fossil constraints in relaxed clock estimation of age and
substitution rates. Am. J. Bot., 100, 556-573.
9.
Posada D. (2008). jModelTest: phylogenetic model averaging. Mol. Biol. Evol., 25,
1253-1256.
10.
Sauquet H., Ho S.Y.W., Gandolfo M.A., Jordan G.J., Wilf P., Cantrill D.J., Bayly M.J.,
Bromham L., Brown G.K., Carpenter R.J., Lee D.M., Murphy D.J., Sniderman
J.M.K. & Udovicic F. (2012). Testing the impact of calibration on molecular
divergence times using a fossil-rich group: the case of Nothofagus (Fagales).
Syst. Biol., 61, 289-313.
11.
Tamura K., Peterson D., Peterson N., Stecher G., Nei M. & Kumar S. (2011). MEGA5:
molecular evolutionary genetics analysis using maximum likelihood,
evolutionary distance, and maximum parsimony methods. Mol. Biol. Evol., 28,
2731-2739.
6