Content uploaded by Wen Jun
Author content
All content in this area was uploaded by Wen Jun on May 16, 2020
Content may be subject to copyright.
Annals of Botany 125: 937–953, 2020
doi: 10.1093/aob/mcaa011, available online at www.academic.oup.com/aob
© The Author(s) 2020. Published by Oxford University Press on behalf of the Annals of Botany Company.
All rights reserved. For permissions, please e-mail: journals.permissions@oup.com.
Atranscriptome-based study on the phylogeny and evolution of the
taxonomically controversial subfamily Apioideae (Apiaceae)
Jun Wen1,2,†, Yan Yu1,†, Deng-Feng Xie1, Chang Peng1, Qing Liu2, Song-Dong Zhou1 and Xing-Jin He1,*
1Key Laboratory of Bio-Resource and Eco-Environment of Ministry of Education, College of Life Sciences, Sichuan University,
Chengdu 610065, Sichuan, P.R. China and 2Key Laboratory of Mountain Ecological Restoration and Bioresource Utilization,
Chengdu Institute of Biology, Chinese Academy of Sciences, Chengdu 610041, Sichuan, P.R. China
*For correspondence. E-mail xjhe@scu.edu.cn
†These authors contributed equally to this work.
Received: 6 June 2019 Returned for revision: 28 October 2019 Editorial decision: 21 January 2020 Accepted: 28 January 2020
Published electronically 4 February 2020
• Background and Aims A long-standing controversy in the subfamily Apioideae concerns relationships among
the major lineages, which has prevented a comprehensive study of their fruits and evolutionary history. Here we
use single copy genes (SCGs) generated from transcriptome datasets to generate a reliable species tree and explore
the evolutionary history of Apioideae.
• Methods In total, 3351 SCGs were generated from 27 transcriptome datasets and one genome, and further
used for phylogenetic analysis using coalescent-based methods. Fruit morphology and anatomy were studied in
combination with the species tree. Eleven SCGs were screened out for dating analysis with two fossils selected
for calibration.
• Key Results A well-supported species tree was generated with a topology [Chamaesieae, (Bupleureae,
(Pleurospermeae, (Physospermopsis Clade, (Group C, (Group A, Group B)))))] that differed from previous trees.
Daucinae and Torilidinae were not in the tribe Scandiceae and existed as sister groups to the Acronema Clade. Five
branches (I–V) of the species tree showed low quartet support but strong local posterior probabilities. Dating analysis
suggested that Apioideae originated around 56.64Mya (95% highest posterior density interval, 45.18–73.53Mya).
• Conclusions This study resolves a controversial phylogenetic relationship in Apioideae based on 3351 SCGs
and coalescent-based species tree estimation methods. Gene trees that contributed to the species tree may under-
going rapid evolutionary divergence and incomplete lineage sorting. Fruits of Apioideae might have evolved in
two directions, anemochorous and hydrochorous, with epizoochorous as a derived mode. Molecular and morpho-
logical evidence suggests that Daucinae and Torilidinae should be restored to the tribe level. Our results provide
new insights into the morphological evolution of this subfamily, which may contribute to a better understanding of
species diversication in Apioideae. Molecular dating analysis suggests that uplift of the Qinghai–Tibetan Plateau
(QTP) and climate changes probably drove rapid speciation and diversication of Apioideae in the QTP region.
Key words: Apioideae, transcriptome, single copy genes, coalescent-based method, species tree, phylogeny, fruit,
evolutionary history.
INTODUCTION
The Apiaceae (Umbelliferae) is a large and readily identiable
family of owering plants, consisting of c. 430 genera and 3780
species that are treated as four subfamilies: Mackinlayoideae,
Azorelloideae, Saniculoideae and Apioideae (Plunkett etal.,
2004; Calviño etal., 2016). Although the subfamilies are con-
sidered to be monophyletic, many tribes and subtribes trad-
itionally recognized within them are not all monophyletic,
especially Apioideae. This subfamily includes c. 404 genera
and 2827–2935 recognized species, making it the largest and
most taxonomically complex group of Apiaceae (Pimenov
and Leonov, 1993). Since the rst monograph, Plantarum
Umbelliferarum Distributio Nova, of Umbelliferae by Morison
(1672), many taxonomic studies of Apiaceae have appeared
(Linnaeus, 1753; Koch, 1824; Bentham, 1867; Drude, 1898;
Kozo-Poljansky, 1916; Cerceau-Larrival, 1962; Pimenov and
Leonov, 1993), and which rst identied Apioideae based
on fruit characteristics and further divided the subfamily into
eight tribes and ten subtribes (Drude, 1898). However, it is
becoming more difcult to handle the phylogenetic relation-
ships within Apioideae based on traditional classication sys-
tems (Drude, 1898; Heywood, 1986; Shneyer etal.,1992) with
more and more species being discovered. With the develop-
ment of molecular biology, an increasing number of molecular
markers have been used for phylogenetic analysis of Apioideae,
including: nuclear ribosomal DNA (nrDNA) internal tran-
scribed spacer (ITS) sequences (Downie and Katz-Downie,
1996a; Downie etal., 1998, 2001, 2010; Katz-Downie etal.,
1999; Lee etal., 2001; Spalik and Downie, 2001, 2007; Spalik
etal., 2004, 2010; Calviño etal., 2006; Ajani etal., 2008; Zhou
etal., 2008; Banasiak etal., 2013); chloroplast DNA (cpDNA)
genes rbcL (Plunkett etal., 1996a, 1997) and matK (Plunkett
etal., 1996b, 1997); intron sequences rpoC1 (Downie etal.,
Downloaded from https://academic.oup.com/aob/article-abstract/125/6/937/5722139 by Sichuan University user on 14 May 2020
Wen etal. — Transcriptome-based study of the phylogeny and evolution of Apioideae
938
1996b, 1998, 2000), rpl16 (Downie etal., 2000; Zhou etal.,
2009) and rps16 (Zhou etal., 2009; Calviño etal., 2016); and
cpDNA restriction sites (Plunkett and Downie, 1999; Lee and
Downie, 2000). In total, 41 major clades within Apioideae have
been identied, of which 21 have been recognized at tribe or
subtribe rank, and many morphology-based genera have been
split and dispersed among the major clades of the new div-
ision based on molecular analysis. However, the relationships
inferred from different molecular markers among the major
clades were ambiguous and controversial (Zhou etal., 2009;
Downie etal., 2010; Spalik etal., 2010; Banasiak etal., 2013;
Calviño etal., 2016), and some clades were even revealed as
not being monophyletic. Thus, a new classication systemof
Apioideae is urgently needed.
The conict in phylogenetic relationships among major lin-
eages referred in previous studies suggests that hybridization or
incomplete lineage sorting (ILS) may be common in Apioideae.
Nevertheless, regardless of the molecular data used, the phylo-
genetic resolution at some nodes is not generally accepted. As a
result, there is a need for genes that can clearly resolve the rela-
tionships among the major clades at credible resolution. Some
recent studies using single copy genes (SCGs) to determine the
phylogeny of plant and animal taxa have shown that the resolu-
tions of the phylogenetic trees have improved when compared
to previous studies (Dunn etal., 2008; Duarte etal., 2010;
Hochbach etal., 2015; Teasdale etal., 2016; Leebens-Mack
etal., 2019). SCGs are common across angiosperm genomes
and are involved in essential housekeeping functions that are
highly conserved across all eukaryotes (De Smet etal., 2013).
Shared SCGs in owering plants are in the unique position of
having the closest resemblance to strictly orthologous genes in
their genomes, and were considered to contain a greater pro-
portion of phylogenetically informative sites than commonly
used protein-coding sequences from the plastid or mitochon-
drial genomes (Duarte etal., 2010; De Smet etal., 2013).
Additionally, SCGs are not subject to concerted evolution as
compared with rDNA sequences, and thus allow homologous
comparisons (Small etal., 2004; Wu etal., 2006; Li etal., 2008),
and their stable copy number and easily assessed orthology
have made them suitable for phylogenetic analysis (Yuan etal.,
2009). Moreover, SCGs obtained from the transcriptome are
coding sequences, and this kind of sequences were suggested
to be less subject to ILS than non-coding sequences due to fre-
quent selective sweeps, which tend to remove ILS (Scally etal.,
2012). Hence, SCGs may be more suitable than other molecular
markers for the phylogenetic analysis of Apioideae.
Deep phylogenetic relationships cannot be reconstructed ac-
curately with small datasets, especially within groups that have
experienced rapid evolutionary divergence, ILS and/or reticulate
evolution (Dunn etal., 2008; Jian etal., 2008; Han etal., 2014;
Zeng etal., 2014). Phylogenetic estimation of species trees based
on genomic datasets might resolve branches that were poorly sup-
ported based on smaller datasets (Rokas etal., 2003; Dunn etal.,
2008; Hochbach etal., 2015; Teasd al e etal., 2016; Leebens-Mack
etal., 2019). Thus, in this study, we extracted thousands (3351)
of shared SCGs from throughout transcriptome datasets and
used them to reconstruct the phylogeny of Apioideae. However,
it is a challenge to incorporate such large amounts of data into
the inference of a species tree, because conicting genealogical
histories often exist in different genes throughout the genome
(Degnan and Rosenberg, 2009). Many phylogenetic studies
have been based primarily on concatenation-based methods to
analyse multiple genes, on the assumption that all genes have
the same or similar phylogenies (William and Ballard, 1996; de
Queiroz and Gatesy, 2007). However, this revealed that simple
concatenation methods could not yield an accurate result for the
phylogenetic relationships of an organism (returning incorrect
trees with high condence) in the presence of gene tree hetero-
geneity (Mossel and Vigoda, 2005; Kubatko and Degnan, 2007;
Song etal., 2012), and this method is computationally intensive
for large datasets and can be statistically inconsistent under the
multi-species coalescent (MSC) model (Roch and Steel, 2014).
Numerous methods have been developed to address this short-
coming, and commonly modelled by the MSC model, which has
emerged as the standard method for reconstructing species trees
in the presence of gene tree discordance due to ILS (Maddison,
1997; Degnan and Rosenberg, 2006; Zhong etal., 2013). Two
main categories of model are used. One is Bayesian coalescent-
based model, which always co-estimate gene trees and the spe-
cies tree, such as *BEAST (Heled and Drummond, 2010), BEST
(Edwards etal., 2007; Ronquist etal., 2012), BPP (Ya n g a n d
Rannala, 2014; Flouri T etal., 2018) and revBayes (Hohna
etal., 2016). Another group is the most scalable family of MSC-
based methods (Sayyari and Mirarab, 2016), which are based
on a two-step process in which gene trees are rst estimated
independently for each gene and then summarized to build the
species tree using a summary method. Many such methods are
widely used, including MP-EST (Liu etal., 2010), STAR (Liu
etal., 2009), NJst (Liu and Yu, 2011) and ASTRAL/ASTRAL-II
(Mirarab etal., 2014; Mirarab and Warnow, 2015; Sayyari and
Mirarab, 2016). MP-EST can estimate branch lengths in coales-
cent units; ASTRAL-II estimates species tree topologies, branch
lengths and local posterior probabilities; and the remainder only
infer the topology.
A reliable species tree often has biological signicance and
can reect the evolutionary trend of species. The evolutionary
process of speciation includes not only changes to the genetic
material, but also changes to the morphological, physiological
and behavioural traits of species. Once a robust species tree of
Apioideae is available, the true morphological evolution of this
subfamily can be explored. Researchers have previously tried to
match morphological evolution with the molecular framework in
the subfamily (Jiménez-Mejías and Vargas, 2015; Wojewódzka
etal., 2019). The Apioideae exhibit considerable morphological
diversity, especially in fruit characters, which is a model trait
that has received signicant attention in studies of the genetic
basis of morphological change (Xiang etal., 2017), and they
have long played a key role in the classication of Apioideae
(Jiménez-Mejías and Vargas, 2015; Zakharova etal., 2016; Liu
and Downie, 2017; Ly skov etal., 2017; Woj ew ódzka etal.,
2019). However, the absence of unique characters for most of
the major clades and the incongruence between phylogenies
make it difcult to describe synapomorphies supporting natural
groups. This has led to the interpretation that morphological
characters have been highly labile during the differentiation of
Apioideae (Jiménez-Mejías and Vargas, 2015). It was also of
note that diversity in one character did not imply diversity in
another among Apioideae, and that one character and its states
may vary tremendously, while another was xed and totally in-
variable (Liu etal., 2016). This indicated that diversication in
Downloaded from https://academic.oup.com/aob/article-abstract/125/6/937/5722139 by Sichuan University user on 14 May 2020
Wen etal. — Transcriptome-based study of the phylogeny and evolution of Apioideae 939
Apioideae has not been stepwise or predictable, but random, and
also not gradual but abrupt. Although morphological studies,
mostly of fruit morphology, have been carried out on many taxa
of Apioideae (mainly focusing on the rank of genera or tribe),
there has been no comprehensive study of the whole subfamily.
The aim of the current study was to resolve the long-standing
controversy among the major clades of Apioideae and recon-
struct a robust phylogenetic relationship based on thousands
(3351) of SCGs from transcriptome data of 27 species repre-
senting major lineages of Apioideae. This is the rst attempt
to estimate a species tree of Apioideae through thousands of
genes based on the coalescent-based method. We integrate mo-
lecular analyses with other data (e.g. fruit morphological traits)
to explore the evolution of Apioideae and propose a hypothesis
regarding the complex evolution of fruits. Specically, we at-
tempt to: (1) investigate the phylogenetic relationships among
the major lineages of Apioideae using a phylotranscriptomic
approach, (2) compare and explain the discordance and agree-
ment between the current species tree topology and those pro-
posed in previous studies, (3) explore the possible evolutionary
histories of fruit types in Apioideae, and (4) estimate the diver-
gence time of this group based on SCGs, and roughly infer its
possible biogeographical history.
MATERIAL AND METHODS
We have generated a simple analytical process in this study,
which is shown in a Flowchart (Fig.1), and the specic oper-
ations are as follows.
Sampling and data retrieval
Our selected taxa covered from the basal to the distal lineages
of Apiaceae subfamily Apioideae (Banasiak etal., 2013), and
representative taxa of the 17 major clades were selected based
on the framework of Downie etal. (2010). Atotal of 28 species
were sampled including one outgroup, Sanicula orthacantha var.
stolonifera, which belongs to Saniculoideae (Table 1 ), a subfamily
that is sister to Apioideae (Chandler and Plunkett, 2004; Plunkett
etal., 2004; Calviño etal., 2006, 2016). All the transcriptome
data were newly generated except for Daucus carota, for which
a whole genome sequence has existed (Iorizzo etal., 2016).
The genomic data of D.carota were obtained from Phytozome
(http://www.phytozome.net/). Due to difculties with transcrip-
tome sampling, all the materials were collected from China. All
material was taken from young leaves and frozen in liquid ni-
trogen. Collection details are given in Supplementary Data Table
S1. For the transcriptome datasets, total RNA was extracted
from young leaves, and library preparation and sequencing were
outsourced to Biomarker Technologies (Beijing, China) and
Novogene (Beijing, China). Transcriptomes were sequenced on
Illumina series device platforms generating 150-bp paired-end
reads. We tested the quality of the data and calculated the per-
centage ofGC content, N50 of contigs(bp) and average contig
lengths(bp) (Tab l e 1). Clean reads were assembled using Trinity
with default settings. Raw reads of the 27 transcriptome datasets
have been submitted to the GenBank at NCBI, with accession
numbers SRR8863732–SRR8863758.
Identication of sharedSCGs
To minimize the impact of missing data on the ability to re-
solve phylogenetic relationships condently, we included only
genes that did not have any taxa missing for further phylogenetic
analyses. First, we extracted SCGs of D.carota (Daucinae) from
its genome. During this process, the protein data of D.carota
were compared against itself using the BLASTN program, with
an e-value cutoff of 1e-20. SCGs were then extracted via our
in-house script I.We then compared these SCGs against the 27
newly assembled transcriptome datasets using BLAST with an
e-value cutoff of 1e-20 and predicted the shared SCGs using
OrthoMCL 2.0 (Fischer etal., 2011); the workow is shown in
script II. Additionally, any suggested deleted sequences during
implementation of TrimAl v1.4.1(Capella-Gutierrez etal.,
2009) were removed from the matrix text le and returned for
a second lter. We used the same method to screen out shared
SCGs among Group Aand Group B respectively.
Phylogenetic reconstruction and species tree estimation
TrimAl was used to adjust all the shared SCGs, and suit-
able genes were screened out for phylogenetic reconstruction.
Within the two species tree estimation methods mentioned
above, Bayesian coalescent-based methods readily provide
support but are limited to a small number of species and loci,
and are computationally intensive (Rannala and Yang, 2017);
by contrast, the summary methods are feasible for genome-
scale datasets and are statistically consistent. After comparing
these two MSC-based methods, we adopted the statistically
consistent summary methods as the optimum choice to es-
timate the species tree of the subfamily Apioideae based on
such a large dataset (3351 gene trees). Among the MSC-based
summary methods, ASTRAL-II (Mirarab and Warnow, 2015;
Sayyari and Mirarab, 2016) has substantial advantages over
other coalescent-based methods: it is faster, can analyse much
larger datasets (up to 1000 species and 1000 genes) and has
substantially better accuracy (Mirarab and Warnow, 2015). In
addition, it has been used for many studies based on genome-
scale datasets (Cloutier etal., 2019; Leebens-Mack etal., 2019;
Allio etal., 2020). Generally, we estimated the species tree
using a two-step approach. In the rst step, we used RAxML
v8 (Stamatakis, 2014) to estimate single-gene trees (rooted and
unrooted), where the GTR substitution model and GAMMA-
distributed site rates were applied, and support was assessed
with 100 bootstrap replicates. In the second step, given the large
datasets and the accuracy of species tree estimation (substantial
advantages), we generated a species tree based on these single-
gene trees under the MSC model preferentially implemented
in ASTRAL 5.6.3 (Mirarab and Warnow, 2015; Sayyari and
Mirarab, 2016), which estimates species trees from unrooted
gene trees, and maximizes the number of quartet trees shared
between the gene trees and the species tree. We estimated
branch lengths for internal branches (not terminal branches)
in coalescent units, which is a direct measure of the amount
of discordance in the gene trees, and computed branch support
values (which measure support for a quadripartition rather than
bipartition). We also computed local posterior probabilities
Downloaded from https://academic.oup.com/aob/article-abstract/125/6/937/5722139 by Sichuan University user on 14 May 2020
Wen etal. — Transcriptome-based study of the phylogeny and evolution of Apioideae
940
based on gene tree quartet frequencies, which is a reliable
measure of accuracy that has very high precision and improved
recall compared with multi-locus bootstrapping (Sayyari and
Mirarab, 2016). To provide extra per branch information, we
used ‘-t 4’ and ‘-t 8’ to annotate the species tree for the main
topology and the rst and second alternatives, which calculate
quartet support (q) and local posterior probabilities (pp) for
each branch, respectively.
We also conducted MP-EST and STAR analyses to reduce
the chance of missing any true positives (Mao etal., 2019) in
the context of ASTRAL-II. MP-EST uses a pseudo-likelihood
function of the species tree and estimates branch lengths in co-
alescent units (Liu etal., 2010), whereas STAR uses average
ranks of coalescences (Liu etal., 2009); both of these methods
generate bootstrap support values using non-parametric boot-
strap techniques. Rooted bootstrapped trees were then up-
loaded to ‘The Species Tree Analysis Webserver’ STRAW
(Shaw etal., 2013) to estimate the species tree using STAR
and MP-EST. All three methods are based on summary statis-
tics calculated across all gene trees, with the effect that a small
BLASTN Tr inity assembly
27 transcriptomes
Clean reads
Daucus carata
genomic data
8886 Daucus carata
single copy genes 27 XX.fasta
BLAST and MCL
3369 Shared
single copy genes
3351 Shared
single copy genes
3351 Gene
trees
NJ tree (a tree-clade) Species tree Divergence
time estimation
Bt branch
1184 gene trees
Ot branch
1260 gene trees
Rt branch
405 gene trees
Pt branch
502 gene trees
Bt
Species tree
Ot
Species tree
Rt
Species tree
Pt
Species tree
11 genes
Cluster
ASTRAL concatenated
Tr imAL and then discard
genes not suitable
Alignment and TrimAL
RAxML
MP-EST
STAR
ASTRAL
MP-EST
STAR
ASTRAL
MP-EST
STAR
ASTRAL
MP-EST
STAR
ASTRAL
MP-EST
STAR
Check out trees with
similar topology to the
species tree
Fig. 1. Flowchart of gene selection and phylogeny construction.
Downloaded from https://academic.oup.com/aob/article-abstract/125/6/937/5722139 by Sichuan University user on 14 May 2020
Wen etal. — Transcriptome-based study of the phylogeny and evolution of Apioideae 941
number of genes that signicantly deviate from the coalescent
model will have relatively little effect on the ability to infer the
species tree accurately.
These single-gene trees were further analysed to test the dif-
ferent evolutionary history of the genes. They were clustered into
a tree-clade upon neighbour-joining (NJ) analysis. To calculate
distances in rooted trees, we considered that the clade near the
root is more important than that near the tip of the tree when com-
paring two rooted trees. Therefore, we developed the normalized
sub-node (SN) distance, modied from the normalized RF dis-
tance (Robinson and Foulds, 1981) as follows.
For each clade Ck in tree T, let |Ck| be the number of species
in Ck, and L(Ck) be the levels (or steps?) from the root of the
tree to the root of Ck.
(T1∩T2)
represents the clades shared by
the twotrees:
w(T1,T2)=
!
Ck∈T1
|Ck|/L(Ck)
+!
Ck∈T2
|Ck|/L(Ck)
−2∗
!
C
k∈
(T
1∩
T
2
)
|Ck|/L(Ck
)
The normalized SN distance is derived by dividing w(T1, T2)
by the maximal possible distance w(T1, T2). The normalized
SN distance has been implemented in RASP 4.2 (Yu etal.,
2020). With the distance matrix, an NJ tree was built using
the neighbor program implemented in the PHYLIP package
3.698 (http://evolution.genetics.washington.edu/phylip.html).
Topologically similar trees were gathered, forming a branch
and suggesting the similar evolutionary history of these cor-
responding genes. We generated species trees of the different
branches of the NJ tree with the methods mentioned above, to
compare and explain the discordance and agreement between
topologies of these trees. We performed this analysis to explore
the impact of different datasets for species tree estimation, and
further to extract gene trees thatmost similar to the species tree,
the corresponding genesof which are suitable for further phylo-
genetic analyses.
Morphological character analysis
A total of 26fruit samples were collected from the eld, ex-
cept for Ligusticum jeholense (Sinodielsia Clade, provided by
Junpei Chen). We were unable to collect fruits of Oenanthe
javancia (Oenantheae) and S.orthacantha var. stolonifera
(Saniculoideae, outgroup), and thus fruits of the closely re-
lated species O.hookeri and S.lamelligera were used. We
did not obtain fruits of Aegopodium podagraria (Careae) or
Pternopetalum vulgare (Acronema Clade), or fruits of their
closely related species. The external characteristics of mature
fruits were examined using a stereomicroscope (Nikon SMZ25)
and measured using MAML (Altınordu etal., 2016). The fruits
were placed on the end of the corresponding branches of the
species tree with their real size scaled to the same percentage.
For anatomical studies, the fruits were embedded in parafn and
cut at the midpoint, and dissections were stained with safranin–
fast green (a few were hand-cut and not stained). Drawings
Table 1. Information on the 27 transcriptome datasets generated in thisstudy
Taxon Subfamily Major clade Tissue No. of
clean reads
Size of
data
Percentage
of GC
content
Contig
N50
(bp)
Average
contig length
(bp)
Ostericum grosseserratum Apioideae Acronema Clade Leaf 26428048 2.115G 39.87 1645 1011.22
Pternopetalum trichomanifolium Apioideae Acronema Clade Leaf 31501995 2.555G 39.49 1578 1011.12
Pternopetalum vulgare Apioideae Acronema Clade Leaf 32721054 2.715G 39.9 1513 978.1
Foeniculum vulgare Apioideae Apieae Leaf 35053593 2.975G 39.27 1866 1158.27
Bupleurum chinense Apioideae Bupleureae Leaf 26220538 2.29G 39.78 1181 761.08
Aegopodium podagraria Apioideae Careae Leaf 28389906 2.295G 39.35 1487 942.51
Chamaesium paradoxum Apioideae Chamaesieae Leaf 32327161 2.62G 39.6 1660 1039.5
Coriandrum sativum Apioideae Coriandreae Leaf 34531629 2.895G 39.74 1823 1120.88
Cryptotaenia japonica Apioideae Oenantheae Leaf 30040381 2.46G 39.34 2001 1228.02
Oenanthe javanica Apioideae Oenantheae Leaf 30500455 2.575G 39.34 1429 940.99
Oenanthe thomsonii Apioideae Oenantheae Leaf 27399783 2.295G 39.38 1515 989.51
Haplosphaera phaea Apioideae Physospermopsis Clade Leaf 32082500 2.62G 39.31 1835 1174.28
Nothosmyrnium japonicum Apioideae Pimpinelleae Leaf 29104885 2.575G 40.32 1290 785.91
Pimpinella diversifolia Apioideae Pimpinelleae Leaf 32721054 2.915G 40.16 1440 915.6
Hymenidium davidii Apioideae Pleurospermeae Leaf 25471955 2.08G 39.05 1660 1032.29
Cyclospermum leptophyllum Apioideae Pyramidoptereae Leaf 32146952 2.675G 39.81 1897 1160.14
Anthriscus sylvestris Apioideae Scandiceae
(Scandicinae)
Leaf 30821940 2.56G 39.81 1760 1090.06
Torilis scabra Apioideae Scandiceae (Torilidinae) Leaf 30209020 2.545G 39.61 1609 1036.19
Angelica acutiloba Apioideae Selineae Leaf 26317221 2.31G 40.54 1564 951.05
Angelica decursiva Apioideae Selineae Leaf 29201429 2.5G 40.77 1243 803.38
Cnidium monnieri Apioideae Selineae Leaf 35637090 3.02G 39.4 1690 1017.61
Peucedanum japonicum Apioideae Selineae Leaf 28900668 2.34G 39.46 1618 1023.49
Saposhnikovia divaricata Apioideae Selineae Leaf 26276265 2.12G 39.66 1690 1074.2
Ligusticum jeholense Apioideae Sinodielsia Clade Leaf 23800507 1.91G 39.16 1741 1102.55
Heracleum candicans Apioideae Tordylieae (Tordyliinae) Leaf 30756825 2.385G 40.23 1615 987.75
Pastinaca sativa Apioideae Tordylieae (Tordyliinae) Leaf 30658501 2.31G 40.12 1789 1115.59
Sanicula orthacantha var.
stolonifera
Saniculoideae Leaf 33859594 2.885G 39.45 1778 1116.56
Downloaded from https://academic.oup.com/aob/article-abstract/125/6/937/5722139 by Sichuan University user on 14 May 2020
Wen etal. — Transcriptome-based study of the phylogeny and evolution of Apioideae
942
were made based on photographs taken using a Nikon SMZ25
photograph system, and displayed on the corresponding branch
of the species tree. We were unable to obtain fruit anatomical of
Angelica decursiva and A. acutiloba, thus we provided another
typical fruit anatomical of Angelica (A. dahurica) as a comple-
ment. To evaluate the evolution of fruit characters, we mapped
some onto the nal species trees using the maximum parsimony
method with Mesquite 3.6 (Maddison and Maddison, 2018).
Two key characters were selected, namely the number of devel-
oped fruit ribs and the type of commissural primary ribs (mar-
ginal ribs), and character scores for each taxon were based on
our own observations.
Molecular dating analysis
Previous estimated ages of Apiaceae differ greatly, ran-
ging from 29Mya to 87.4 (76–99)Mya (Bremer etal.,
2004; Bell etal., 2010; Nicolas and Plunkett, 2014; Calviño
etal., 2016), which inevitably affected the estimated diver-
gence time of the subfamily Apioideae. The most recent mo-
lecular dating analyses of Apioideae were made based on the
nrDNA ITS sequence (Banasiak etal., 2013) and cpDNA
rps16 intron (Calviño etal., 2016). In the present study,
gene trees with similar topologies, broadly consistent with
the nal species tree, were screened out by script II, and the
corresponding genes were extracted. We then concatenated
these genes into a large matrix to estimate the divergence
time. BEAST v1.8.4 (Drummond etal., 2012) was used to
estimate the topology, substitution rates and node ages sim-
ultaneously by using a Bayesian Markov chain Monte Carlo
(MCMC) approach. The nucleotide substitution model was
chosen using Modeltest 3.7 (Posada and Crandall, 1998).
We assumed that substitution rates evolved under the un-
correlated lognormal (UCLN) model and the general time
reversible (GTR) model for nucleotide substitution. AYule
prior set was used to estimate divergence times and the cor-
responding credibility intervals. Independent replications
each with 108 generations were run with sampling every 104
generations. The stationarity of the chains and convergence
of two runs was monitored by Tracer, with the effective
sample size of all parameters >200. The chronogram with
nodal heights and 95% highest posterior density intervals
(95% HPDs) was generated with TREEANNOTATOR, with
the rst 1000 trees being discarded as burn-in. Two calibra-
tion points from fossils were used to determine species node
priors – we mainly used the pollen fossil adopted by Banasiak
etal. (2013). The rst calibration point was placed at the
stem node of Bupleureae (Gruas-Cavagnetto and Cerceau-
Larrival, 1984) and was constrained to a lognormal distribu-
tion with a lower bound (offset) of 33.90Mya (the end of the
Priabonian). We also extended the strategy used by Banasiak
etal. (2013)—adjusted the mean to set the upper bound of
the distribution (95% quantile) to 58.7Mya (corresponding
to the beginning of the Thanetian) instead of choosing ar-
bitrary parameters. The second point with a similar con-
straint was imposed for the stem node of Pleurospermeae
(Gruas-Cavagnetto and Cerceau-Larrival, 1984), and
the upper bound was constrained to the beginning of the
Ypresian (55.8Mya). We abandoned the third fossil given by
Banasiak etal. (2013), which was suggested to be attributed
to a monophyletic group comprising Scandiceae, Smyrnieae,
Aciphylleae and the Acronema clade, because (1) informa-
tion regarding this fossil has not been published already; (2)
the study by Calviño etal. (2016) showed that Scandiceae,
Smyrnieae and Aciphylleae could not be used as a monophy-
letic group, which made it is difcult to choose a reasonable
position for calibration. This suggests that information re-
garding this fossil is not reliable, especially given the small
number of species we used. Arbitrary values of 0.5 for the
standard deviation were chosen; all information relating to
the two calibration points is given in Table2.
RESULTS
Sequencing of transcriptome datasets and identication of
sharedSCGs
We sequenced 27 new transcriptome datasets covering most
of the lineages of Apioideae, especially lineages lacking
sequenced genomes. The lineages we selected were distrib-
uted uniformly in the phylogenetic trees of Apioideae re-
constructed by Banasiak etal. (2013) based on ITS, eight of
which belong to the apioid superclade (Downie etal., 2010).
The transcriptome datasets were large enough and contained
sufcient information for subsequent analysis, and the
quality of these datasets were good as revealed by percentage
of GC content, the N50 of contigs and average contig lengths
(Table1). The percentage ofGC content varies slightly
(ranging from 39.05 to 40.77%), and the average length
of contigs varied from 761.08 to 1228.02bp, along with
N50 values ranging from 1181 to 2001bp. We performed
a series of analyses on the transcriptome datasets and gen-
omic data. Atotal of 8886 SCGs were screened out from the
D.carota genome initially, 3369 of which were identied
as shared SCGs through BLAST and MCL analyses. Within
these shared SCGs, 18 were ltered out through TrimAl pro-
cessing as being unsuitable for phylogenetic tree construc-
tion. As a result, we nally identied 3351 shared SCGs for
the phylogenetic reconstruction of Apioideae. Subsequently,
we screened out 11 shared SCGs for divergence time estima-
tion, and the nal alignments of these 11 concatenated SCGs
reached 19855bp.
Table 2. Temporal constraints used to estimate divergence times for Apioideae (Mya); the standard deviation of the lognormal distri-
bution was set to 0.5 for all calibrationpoints
Stem node (fossil reference) Offset Mean Median 95% quantile
Bupleureae (Gruas-Cavagnetto and Cerceau-Larrival, 1984) 33.90 2.389 44.80 58.71
Pleurospermeae (Gruas-Cavagnetto and Cerceau-Larrival, 1984) 33.90 2.264 43.52 55.80
Downloaded from https://academic.oup.com/aob/article-abstract/125/6/937/5722139 by Sichuan University user on 14 May 2020
Wen etal. — Transcriptome-based study of the phylogeny and evolution of Apioideae 943
Phylogenetic relationships among major lineages of Apioideae
A total of 3351 rooted/unrooted gene trees were obtained
through RAxML, bootstrapped trees were analysed under
all three methods and an extra analysis was implemented in
ASTRAL-II with the best maximum-likelihood (ML) trees.
The same topology was retrieved regardless of the bifurcate
tree-building method used (including MP-EST, STAR and
ASTRAL-II) (Fig.2). All tree-building methods lent strong
branch support values for every branch of the species trees, ex-
cept branch IV, which shows high branch support values in STAR
(100) and ASTRAL-II (97, 1.0) but a moderate MP-EST branch
support value (67). Quartet support analysis in ASTRAL-II sug-
gested that ve branches (I–V) have support values <50% for the
main topology (Fig.2; Supplementary Data Fig. S1), indicating
a considerable gene tree conict around these branches (Sayyari
and Mirarab, 2016). However, the extremely high local posterior
probabilities (pp1=1, pp2=0, pp3=0) for each branch suggest
a very high precision of our speciestree.
Broadly, our species tree indicates that all of the apioid
superclade groups are strongly supported, forming Group
A.Particular branches in this group with almost the same
quartet support for three topologies (the main topology, the
rst alternative,and the second alternative) are III and IV, as-
sociated with the Sinodielsia Clade (L.jeholense) and Apieae
(Fo en i c ul u m v u lg a re ), with the following quartet support values:
III (q1=0.37, q2=0.32, q3=0.31) and IV (q1=0.37, q2=0.29,
q3=0.34) (Fig.2; Supplementary Data Fig. S1). Additionally,
branch lengths for these two branches are extremely short.
Branch V, which links the internal branches in Selineae, is sup-
ported by only 48% of the gene trees in the main topology.
To explore the relationships among major lineages of Group
A, we performed gene resampling analysis. Two taxa were
added in this study, Angelica dahurica and A.laxifoliata, and
3793 shared SCGs were obtained for the phylogenetic recon-
struction of Group A.The species tree of Group Awas esti-
mated using ASTRAL-II based on both ‘bootstrapped trees’
and ‘best ML trees’. The topology of the species tree of Group
A(Supplementary Data Fig. S2) was consistent with this group
within the Apioideae species tree estimated by the former
(Fig.2). However, when compared with the species tree of
Apioideae, the species tree of Group Ashows signicant de-
crease in branch support for branch III, with only moderate (62)
and low support (0.47), and the branch length is signicantly
shortened (Fig. S2). Quartet support analysis suggested that
branches III–V had lower gene tree support than other branches
(Figs S2 and S3).
Furthermore, Group Aand GroupB are strongly sup-
ported forming a sister clade, with Group B comprising
Daucinae (D.carota), Torilidinae (Torilis scabra), Scandicinae
(Anthriscus sylvestris), and the Acronema Clade (including
P. trichomanifolium, P. vulgare and Ostericum grosseserratum)
(Fig.2). Daucinae and Torilidinae are sisters with moderate
quartet support for the main topology (q1=0.64), and subse-
quently forming successive sister clades with the Acronema
Main topology
First alternative
Second alternative
*
0.27 0.48
0.25
0.32 0.37
0.31
0.34
0.37
0.49
0.3
0.22
0.23
0.47
0.3
0.29
*
*
*
*
*
*
*
*
*
*
*
**
*
**
**
*
*
*
*
*
IV III
II
I
V
64/100/97/1.0
Selineae
Peucedanum japonicum
Saposhnikovia divaricata
Cnidium monnieri
Angelica acutiloba
Angelica decursiva
Coriandrum sativum
Pastinaca sativa
Heracleum candicans
Foeniculum vulgare
Ligusticum jeholense
Nothosmyrnium japonicum
Pimpinella diversifolia
Cyclospermum leptophyllum
Aegopodium podagraria
Pternopetalum trichomanifolium
Pternopetalum vulgare
Ostericum grosseserratum
Torilis scabra
Daucus carota
Anthriscus sylvestris
Oenanthe javanica
Oenanthe thomsonii
Cryptotaenia japonica
Haplosphaera phaea
Hymenidium davidii
Bupleurum chinense
Chamaesium paradoxum
Sanicula orthacantha var. stolonifera
Coriandreae
Tordylieae
Apieae
Pimpinelleae
Pyramidoptereae
Careae
To rilidinae
Daucinae
Scandicinae
Pleurospermeae
Physospermopsis Clade
Bupleureae
Chamaesieae
Outgroup
Sinodielsia Clade
Acronema Clade
Oenantheae
Apioideae
Saniculoideae
2.0
Group B
Group C
Group A
D
Fig. 2. Species tree generated using ASTRAL-II based on 3351 single copy gene trees. Numbers or asterisks above branches are branch support values for
MP-EST, STAR and ASTRAL-II with bootstrapping analyses and local posterior probabilities for ASTRAL-II with best maximum-likelihood trees, respectively,
with asterisks denoting maximum support in all four analyses. ASTRAL-II measures branch lengths in coalescent units (the scale bar shown corresponds to two
coalescent units) for internal branches and not terminal branches (branch lengths of terminal branches are therefore arbitrary and meaningless). The pie charts show
respective quartet support for the main topology, and the rst and the second alternative topology.
Downloaded from https://academic.oup.com/aob/article-abstract/125/6/937/5722139 by Sichuan University user on 14 May 2020
Wen etal. — Transcriptome-based study of the phylogeny and evolution of Apioideae
944
Clade (q1= 0.59) and Scandicinae (q1=0.77) (Supplementary
Data Fig. S1). Note that Daucinae, Torilidinae and Scandicinae
were previously assigned to the tribe Scandiceae (Downie
etal., 2010; Banasiak etal. 2013), but herein these three
clades do not constitute a monophyletic group. We therefore
resampled 4257 genes for Group B analysis separately, and
this again revealed that the three subtribes could not form a
monophyletic group (Figs S4 and S5), and the relationships
among these four major clades within Group B were without
any changes. Group C is composed of a natural monophyletic
tribe Oenantheae, containing O.javanica, O.thomsonii and
Cryptotaenia japonica, and the internal branches have very
high quartet support values (Fig. S1). In addition, our species
tree strongly supports Chamaesieae (Chamaesium paradoxum)
as the most basal lineage of the Apioideae we studied here, not
the Bupleureae (Bupleurum chinense). Our species tree shows
that Groups A, B Group C exist as successive sister groups
clustering a stabilized distal branch D [Group C, (Group A,
Group B)], and this branch D along with the Physospermopsis
Clade (Haplosphaera phaea), Pleurospermeae (Hymenidium
davidii), Bupleureae and Chamaesieae then form successive
sister clades.
In further analyses, we clustered all the single-gene trees
according to distance into an NJ tree (Fig.3; Supplementary
Data Fig. S6), which generated four main branches, Rt, Pt, Bt
and Ot, for which the number of gene trees was 405, 502, 1184
and 1260, respectively. The species trees of these four datasets
showed broadly similar topologies except for the basal clades
(Fig.3; Figs S7–S16). The difference between species trees of
the two richer gene trees branches, Bt (Fig. S7) and Ot (Fig.
Group A
Bt
Pt
Ot
Rt
Bt
Pt
Ot
Rt
0.75/* 0.73/*
0.49/*
0.92/*
0.94/*
0.8/*
0.95/*
0.53/*
0.63/*
0.66/*
0.76/*
0.52/*
0.91/*
0.92/*
0.86/*
0.93/*
0.63/*
0.52/*
0.62/*
0.42/*
0.73/*
0.66/*
0.55/*
0.45/*
0.37/(93/0.59)
0.5/*
0.95/* 0.82/*
0.97/*
0.97/*
0.85/*
0.69/*
Group B
Group C
Haplosphaera phaea
Hymenidium davidii
Bupleurum chinense
Chamaesium paradoxum
Sanicula orthacantha var. stolonife ra
Group A
Group B
Group C
Haplosphaera phaea
Hymenidium davidii
Bupleurum chinense
Chamaesium paradoxum
Sanicula orthacantha var. stolonife ra
Group A
Group B
Group C
Haplosphaera phaea
Hymenidium davidii
Bupleurum chinense
Chamaesium paradoxum
Sanicula orthacantha var. stolonife r
a
Group A
Group B
Group C
Haplosphaera phaea
Hymenidium davidii
Bupleurum chinense
Chamaesium paradoxum
Sanicula orthacantha var. stolonife ra
A
B
Fig. 3. A cluster of all the 3351 gene trees and the species trees of the four branches. (A) The neighbor-joining tree of gene trees; all trees gathered into four
branches marked by Rt, Pt, Bt and Ot. (B) The species trees of these four branches. Numbers in the rst column above branches are quartet support values, and the
asterisks and numbers in parentheses above branches are branch support values for ASTRAL-II with bootstrapping analyses and local posterior probabilities for
ASTRAL-II with best maximum-likelihood trees, respectively, with asterisks denoting maximum support in both analyses.
Downloaded from https://academic.oup.com/aob/article-abstract/125/6/937/5722139 by Sichuan University user on 14 May 2020
Wen etal. — Transcriptome-based study of the phylogeny and evolution of Apioideae 945
S10), lies in whether Bupleureae or Chamaesieae is the most
basal clade. The Bt species tree supports Chamaesieae at the
most basal clade, which is consistent with the nal species
tree (i.e. the species tree estimated based on 3351 gene trees),
whereas the Ot species tree has Bupleureae as the most basal
lineage. Within the Rt speciestree (Fig. S12), Chamaesieae is
still the most basal lineage of the subfamilyApioideae, and
Pleurospermeae then separated earlier than Bupleureae. The
Pt species tree (Fig. S14) shares the same topology with the
Bt and nal species trees but with lower quartet support values
for each branch, especially for the most basal internal branch,
and this branch also has depressed branch support (93/0.59).
We used the ‘-q’ option in ASTRAL-II to score the species
trees of the four branches and the nal species tree, with the
normalized quartet score used as a measure of discordance for
the gene trees: the higher number, the less discordant are the
gene trees (Sayyari and Mirarab, 2016). In our analyses, the
normalized quartet score of the Bt tree is the highest (0.859)
and Pt is the lowest (0.650), while those of Ot (0.832) and Rt
(0.818) are slightly lower than Bt (Table S2). This indicates
that the Pt datasets has the highest ILS level, whereas Bt has
the lowest.
Fruit characters of Apioideae
The sampled taxa represent a wide array of the fruit morph-
ologies found in the subfamily. Photographs of fruits and
sections through them are shown in Fig.4. Fruit character
explanations are given in Fig.4A, for which we base all of
the characters described below. The Apioideae display a wide
spectrum of fruit morphotype, varying greatly in fruit size,
number and type of ribs, shapes of fruit transverse sections, etc.
(Fig.4B). Supercial observations of fruit size suggested that
fruits of the basal lineages were general smaller than those at
the tip of the species tree, including Chamaesieae, Bupleureae,
Oenantheae and some taxa of Group A, with these taxa ex-
isting as small herbs; larger fruits appear within tall, robust
Umbellifers, such as Selineae, Tordylieae, the Acronema Clade
(Ostericum), and some basal lineages such as Pleurospermeae
and the Physospermopsis Clade. Anatomical studies indicated
that these fruits may be roughly grouped into three categories
according to rib number and type: fruits with co-developed
primary and secondary ribs, having a total of nine ribs
(Chamaesieae); fruits with unique prominent secondary ribs
(Daucinae and Torilidinae), generally having four ribs; and
fruits of the remaining groups with developed primary ribs and
degradative secondary ribs, these always containing ve ribs
(Fig.4; Supplementary Data Fig. S17). The type of marginal
rib within the third fruit category varies widely from liform
to winged, including liform, subtriangular, narrowly winged
and widely winged (Fig. S17). Fruits with widely winged
marginal ribs are particularly common in tribes Selineae,
Tordylieae, and some members of the Acronema Clade
(Ostericum), and dorsal and lateral primary ribs of these fruits
are always liform or slightly winged (Fig.4). Although fruits
of Pleurospermeae (Hymenidium) and the Physospermopsis
Clade (Haplosphaera) also have winged marginal ribs, the
marginal ribs are only slightly developed compared with the
other primary ribs. This kind of fruit also appears in the basal
lineages of Group A(Apieae and Sinodielsia Clade). Fruits
with subtriangular marginal ribs occur in Group C, and have
seeds that are always covered by corky mesocarp and waxy
pericarp. Fruits are dramatically diverse in Group B, including
fruits with liform marginal ribs (some taxa of the Acronema
Clade, e.g. Pternopetalum), with widely winged marginal ribs
as mentioned above (Ostericum), without any prominent ribs
(Scandicinae, e.g. Anthriscus) and with developed secondary
ribs (Daucinae and Torilidinae). Specialization of fruit ribs is
common in Apioideae, mainly falling into two types: spiny ap-
pendages (sometimes winged) that occur on secondary ribs,
found only in Daucinae and Torilidinae; and specialization
of the primary ribs, especially the marginal ribs, which gen-
erally extend into wings. This specialization occurs in most
fruits of the subfamily Apioideae, especially in Group Aand
GroupB.Fruits of the outgroup, S.lamelligera (Saniculoideae),
without any prominent ribs, are densely covered with irregu-
larly arranged straight bristles.
Divergence time estimates compared with previous studies
The estimated times here from BEAST analyses using the
concatenated dataset of 11 SCGs (Fig.5) were similar to those
based on ITS sequences (Banasiak etal., 2013) with devi-
ations of c. 1–6Myr (Table3), except for the divergence time
of tribe Chamaesieae, which is almost identical to those of
the mean times assessed under a PL (penalized-likelihood)
method based on rps16 sequences (Calviño etal., 2016), with
a difference of only c.1.62 Myr. The most recent common
ancestor (MRCA) of Saniculoideae diverged from that of
Apioideae at 56.64Mya (95% HPD, 45.18–73.53Mya) during
the Late Cretaceous–Eocene, just c. 2.86Myr later than esti-
mated by Calviño etal. (2016) (Fig.5; Table3). This suggests
the fossils we used to estimate the time of divergence in the
Apioideae may be appropriate, contrary to the views of Calviño
etal. (2016). Within the subfamily Apioideae, the MRCAs
of Chamaesieae, Bupleureae and Pleurospermeae were esti-
mated to have diverged from others during the Eocene at c.
49.78, 44.88 and 41.15Mya, respectively. The MRCA of the
Physospermopsis Clade later diverged from the distal branch
D at 33.07Mya (27.44–40.01Mya), the divergence between
Group C and the MRCA of Group A+B occurred 26.39Mya
(20.78–32.13Mya), and Group Aand Group B split from
each other at 24.51Mya (19.32–29.98Mya). Within Group
A, the MRCA of Careae and Pyramidoptereae separated from
the MRCA of the remaining lineages at 21.38Mya (16.8–
26.5Mya). All the major lineages of Group Aoriginated and
diversied between c. 21.24 and 7.99Mya. The phylogeny
position of the Sinodielsia Clade differs from the species
tree; this lineage diverged from the MRCA of Tordylieae and
Apieae at 13.21Mya (9.65–17.07Mya), and Tordylieae di-
verged from Apieae at 12.29Mya (8.84–16.11Mya). The
subtribe Scandicinae, which is located as the most basal
branch of Group B, diverged from the MRCA of the re-
maining lineages at c. 19.44Mya (14.67–24.52Mya); the
separation of subtribes Daucinae and Torilidinae was during
the middle Miocene, c. 12.01Mya (8.21–16.66Mya), and the
sister groups split from the Acronema Clade at c. 14.52Mya
(10.61–19.41Mya).
Downloaded from https://academic.oup.com/aob/article-abstract/125/6/937/5722139 by Sichuan University user on 14 May 2020
Wen etal. — Transcriptome-based study of the phylogeny and evolution of Apioideae
946
DISCUSSION
Rapid evolutionary divergence and inference of phylogenetic
relationships among major lineages of Apioideae
The main aim of this study was to resolve and explain the
long-standing controversy of relationships between the major
lineages in Apioideae. As we have shown, the species trees
constructed using MP-EST, STAR and ASTRAL-II all show
a maximally supported topology [Chamaesieae, (Bupleureae,
(Pleurospermeae, (Physospermopsis Clade, (Group C, (Group
A, Group B)))))], with the positions of some major lineages
differing from previous studies (Zhou etal., 2009; Banasiak
etal., 2013; Calviño etal., 2016). The relationships among the
basal lineages are similar to those inferred from cpDNA data
only (Zhou etal., 2009; Calviño etal., 2016), and our place-
ments of the Sinodielsia Clade and Apieae conicted with
Selineae
Coriandreae
Tordylieae
Apieae
Sinodielsia clade
Pimpinelleae
Careae
Pyramidoptereae
To rilidinae
Daucinae
Scandicinae
Oenantheae
Pleurospermeae
Bupleureae
Chamaesieae
Outgroup Saniculoideae
Apioideae
Physospermopsis clade
Acronema clade
A1 A2 A3 A4 A5 A6 A8 A7 A9 A10 A12 A11 A13 B2 B1 B4 B5
Group B
Group A
Group C
B3 C1 C2 C3 DE FG H
Dorsal primary rib (median)
Lateral primary rib
Vallecular vittae
Commissural primary rib
(marginal)
Vascular bundle
Dorsal secondary rib
Lateral secondary rib
(lateral)
Commissural vittae Endosperm
A B
D
Fig. 4. Morphological variation and evolutionary histories of Apioideae. (A) Descriptions of the fruit characters considered in this study. (B) Morphologies of
fruit appearance displayed at the end of the corresponding branches of the bifurcated tree, in the context of the phylogeny shown in Fig.2, with all fruits scaled
to their true size at the same percentage. Taxa in the purple, red and green boxes form Group A, GroupB and GroupC, respectively. D refers to distal branch
D.A1: Angelica decursiva; A2: A.acutiloba; A3: Peucedanum japonicum; A4: Saposhnikovia divaricata; A5: Cnidium monnieri; A6: Coriandrum sativum; A7:
Pastinaca sativa; A8: Heracleum candicans; A9: Foeniculum vulgare; A10: Ligusticum jeholense; A11: Nothosmyrnium japonicum; A12: Pimpinella diversifolia;
A13: Cyclospermum leptophyllum; B1: Pternopetalum trichomanifolium; B2: Ostericum grosseserratum; B3: Anthriscus sylvestris; B4: Torilis scabra; B5:
Daucus carota; C1: Oenanthe hookeri; C2: O.linearis; C3: Cryptotaenia japonica; D: Haplosphaera phaea; E: Hymenidium davidii; F: Bupleurum chinense; G:
Chamaesium paradoxum; H: Sanicula lamelligera. Adrawing of a transverse section through each fruit is displayed on the corresponding branch of the phylogen-
etic tree. An extra fruit transverse section ‘I’ shown at the intersection of two Angelica species(A1 and A2) belongs to A.dahurica.
Downloaded from https://academic.oup.com/aob/article-abstract/125/6/937/5722139 by Sichuan University user on 14 May 2020
Wen etal. — Transcriptome-based study of the phylogeny and evolution of Apioideae 947
Unrelated to uplift
Related to uplift Related to uplift and
climate changes
2.64 Mya (1.38, 4.26)
5.32 Mya (3.6, 7.25)
11.30 Mya (7.99, 15.01)
21.38 Mya (16.8, 26.5)
15.9 Mya (10.63, 21.24)
24.51 Mya (19.32, 29.98)
26.39 Mya (20.78, 32.13)
12.38 Mya (7.05, 19.18)
33.07 Mya (27.44, 40.01)
41.15 Mya (36.65, 47.8)
44.88 Mya (39.49, 52.41)
49.78 Mya (42.72, 58.7)
56.64 Mya (45.18, 73.53)
19.44 Mya (14.67, 24.52) 12.01 Ma (8.21, 16.66)
2.67 Mya (1.3, 4.57)
4.47 Mya (2.3, 7.44)
8.79 Mya (5.35, 12.62)
14.52 Mya (10.61, 19.41)
Group A
Group B
D
Group C
12.29 Mya (8.84, 16.11)
13.29 Mya (9.65, 17.07)
Selineae
Apioideae
Saniculoideae
Coriandreae
Tordylieae
Apieae
Sinodielsia clade
Pimpinelleae
Careae
Daucinae
To rilidinae
Scandicinae
Oenantheae
Pleurospermeae
Physospermopsis clade
Bupleureae
Chamaesieae
Outgroup
70 60 50
Cretaceous
Palaeocene Eocene
Start of the
India-Asia collision
(East)
End of the
India-Asia collision
(West)
Start of the uplift
of West Kunlun rangeIntense uplift of the Hengduanshan
and of Qiadam basin, formation of
the Yangzi, Mekong and Salween valleys
Progressive uplift
of the Tianshan and
the Himalayas
Start of uplift
(South QTP)
Start of the uplift
of Qiadam basin
Oligocene Miocene Plio. Quat.
40 30 20 10 0Mya
Pyramidoptereae
Acronema clade
Angelica decursiva
Angelica acutiloba
Peucedanum japonicum
Saposhnikovia divaricata
Cnidium monnieri
Coriandrum sativum
Pastinaca sativa
Heracleum candicans
Foeniculum vulgare
Ligusticum jeholense
Pimpinella diversifolia
Nothosmyrnium japonicum
Aegopodium podagraria
Cyclospermum leptophyllum
Pternopetalum vulgare
Pternopetalum trichomanifolium
Ostericum grosseserratum
Daucus carota
Tor ilis scabra
Anthriscus sylvestris
Oenanthe javanica
Oenanthe thomsonii
Cryptotaenia japonica
Haplosphaera phaea
Hymenidium davidii
Bupleurum chinense
Chamaesium paradoxum
Sanicula orthacantha var. stolonifera
Fig. 5. Chronogram presenting estimated divergence times by BEAST using 11 single copy genes and calibrated with Tertiary fossil pollen. Calibration points are marked with black circles. Blue bars and
numbers above represent the 95% highest posterior density (95% HPD) for each node. The scale axis is in millions of years ago (Mya). Below the scale axis shows the geological sequence of events related
to the uplift of the QTP, including a graphical representation of the extent of the uplift through time. The top purple axis is the time period associated with QTP uplift and climate change (with reference to
Faver etal., 2015).
Downloaded from https://academic.oup.com/aob/article-abstract/125/6/937/5722139 by Sichuan University user on 14 May 2020
Wen etal. — Transcriptome-based study of the phylogeny and evolution of Apioideae
948
those inferred from all of the published phylogenies (Zhou
etal., 2009; Banasiak etal., 2013; Calviño etal., 2016). The
very low support value (62/0.47) for branch III in the species
tree of Group A(resampled datasets, Supplementary Data Fig.
S2), and topologies of MP-EST species trees of Bt and Pt (with
weak or moderate support, Figs S8, S15) conict with those
generated by other methods (STAR and ASTRAL-II), and the
low quartet support values of branches III and IV regardless of
the dataset used (Figs S1, S3, S9, S11, S13 and S16) suggest
that the position of the Sinodielsia Clade and Apieae remains to
be conrmed with high condence. Our species tree strongly
supports Cyclospermum letophyllum (Pyramidoptereae) as a
sister group to Careae, the placement of which was variable
in previous analyses based on ITS and rps16 intron (Banasiak
etal., 2013; Calviño etal., 2016). The relationship among lin-
eages in Group B is similar to that of Zhou etal. (2009) based
on cpDNA, although sampling for both studies is incomplete
and additional studies are needed.
Our analyses of the weighted datasets show that the species
tree of the Pt dataset, with very high ILS level, shares the same
topology with the nal species tree, indicating that ASTRAL-II
can reconstruct an accurate species tree in the context of a high
ILS level. However, when the ILS level is low, the weighted
datasets (Ot and Rt) may produce false species trees, indicating
that in this situation the input gene trees have a large impact
on the estimation of the species tree. This explains why we
tend to get conicting species trees based on small datasets,
because such analyses may only yieldparticular trees in par-
ticular cases. Because this subfamily has a large number of
species, when the species number is increased dramatically,
existing species tree estimation methods, even ASTRAL-II,
could not be run in a reasonable time. Thus, although weighted
datasets appear to be unscientic for species trees estimation,
we can use the weighted cluster method to reversely analyse
and screening out a set of genes with similar tree topologies
to the nal species tree; these genes can be applied to the re-
construction of the phylogenetic framework of the subfamily
Apioideae. This study provides a feasible method for the esti-
mation of species trees in the subfamily Apioideae, which will
be further discussed in future publications. Our results show
that the topologies of gene trees for the Bt dataset should be
the most similar to that of the nal species tree, and thus we
speculate that the corresponding genes of this dataset may play
key roles in the whole evolutionary histories of Apioideae. It
would be of value to explore the function of these genes, which
may help to explain how the subfamily Apioideae has evolved.
In our ASTRAL-II analyses, branches I–V received one
high-frequency topology and two lower frequency topologies
(Fig.2), as expected when the conict between gene trees is
caused by ILS, which is always a close companion of rapid
evolutionary divergence (e.g. Maddison, 1997). The incongru-
ence among these branches is unlikely to be explained using
hybridization and introgression during their early evolutionary
history. If the conict between gene trees is caused by hybrid-
ization and introgression, one might expect two major (and
equivalent) frequency gene tree topologies (e.g. if the branch
was a result of hybrid speciation) or some other set of frequen-
cies (e.g. if a sub-set of the genome introgressed at this point).
Thus, the pattern of gene tree topology frequencies we found
above is more consistent with the scenario of ILS than with
hybridization and introgression (Mao etal., 2019). The rela-
tively short branch lengths of these branches additionally indi-
cate high levels of ILS (Mirarab and Warnow, 2015), consistent
with rapid evolutionary divergence.
Fruit dispersal modes and character evolution of Apioideae based
on the speciestree
The diaspores of Apiaceae fruits are usually described based
on their shape and appendages as anemochorous (dispersed by
wind), epizoochorous (carried away on animal fur or feathers),
hydrochorous (oating on water), or without any distinct
adaptations to dispersal as barochorous (gravity-dispersed)
(Wojewódzka etal., 2019). The various specializations of ribs
and pericarp were conducive to seed dispersal (Fuentes and
Vivian-Smith, 2009). In our study, fruits of Group C are clearly
hydrochorous, with the corky mesocarp and waxy pericarp
improving oating capabilities, indicating these fruits may fa-
vour long-distance dispersal by water. In some additional habi-
tats (far from owing water) they were barochorous in order to
maintain the local population.
Fruit appendages facilitating wind dispersal may be of dif-
ferent origins, such as developing from either primary (for most
Table 3. Comparison of estimated divergence times (Mya) of major lineages; time was estimated by BEAST methods based on 11 con-
catenated single copygenes
Node ITS rps16 Single copy genes
Group A(Pyramidoptereae – Selineae) 26.1 (22.14, 30.4) About 30.5 21.38 (16.8, 26.5)
Group B (Scandicinae – Acronema Clade) 25.06 (22, 28.7) About 26.5 19.44 (14.67, 24.52)
Group A– Group B About 31.5 About 33.5 24.51 (19.32, 29.98)
Group C (Oenantheae) 22.92 (19.16, 26.97) About 30.5 12.38 (7.05, 19.18)
(Group A+ Group B) – Group C 32.12 (28.41, 36.20) About 36.5 26.39 (20.78, 32.13)
Physospermopsis Clade 21.7 (14.92, 30.06) ? 33.07 (27.44, 40.01)
Pleurospermeae 38.9 (35.97, 42.76) About 47 41.15 (36.65, 47.8)
Bupleureae 44.51 (39.11, 51.55) About 49.5 44.88 (39.49, 52.41)
Chamaesieae 20.25 (14.05, 27.61) 51.4 (37.2, 68) 49.78 (42.72, 58.7)
Saniculoideae – Apioideae ? About 59.5 56.64 (45.18, 73.53)
Times estimated based on ITS refer to Banasiak etal. (2013) with three fossil calibrations; times estimated based on rps16 under a penalized-likelihood (PL)
method refer to Calviño etal. (2016) with one fossil calibration; times estimated based on single copy genes are from our research with two fossil calibrations.
Ranges correspond to 95% highest posterior density (95% HPD) or approximate bootstrap condence quadratic (ABCq) condence intervals. ‘?’ None available.
Downloaded from https://academic.oup.com/aob/article-abstract/125/6/937/5722139 by Sichuan University user on 14 May 2020
Wen etal. — Transcriptome-based study of the phylogeny and evolution of Apioideae 949
taxa of Apioideae) or secondary ribs (within taxa of Daucinae,
such as Thapsia and Laserpitium) (Weitzel etal., 2014;
Banasiak etal., 2016; Wojewódzka etal., 2019). The only study
on wind dispersal in winged-fruited Umbellifers concluded that
its effectiveness as a mode of dispersal is quite low (Jongejans
and Telenius, 2001). Other studies have suggested that bris-
tles and hooks increase attachment and retention potential on
animal fur (Römermann etal., 2005; Tackenberg etal., 2006)
and promote gene ow among populations (Williams, 1994;
Williams and Guries, 1994); recent research by Wojewódzka
etal. (2019) suggested that epizoochorous fruits were generally
dispersed at higher distances than winged fruits. These studies
all rejected a role of anemochory in dispersal of Apioideae
fruits. However, we reviewed these studies and suggest that the
design of these studies might have ignored the inuence of the
environment on the survival of plants (habitat, wind frequency,
type and population of local vectors, etc.), or were based on
only a particular group (Wojewódzka etal., 2019), which was
not sufcient to represent the fruit evolution characteristics of
the entire subfamily. Our results have shown that the species
with widely winged fruits generally live in high-altitude areas
with an open environment, which would favour long-distance
dispersal by air through ight (Theobald, 1971; Downie etal.,
2002; Spalik etal., 2004; Calviño etal., 2008, 2016; Fernández
etal., 2017a, b) and expand the dispersal distance by occupying
more living space. We also suggest that characters such as fruit
size, degradation of the dorsal and lateral primary ribs, and
compression of the endosperm are an adaptation to this dis-
persal model. Winged fruits have been present throughout the
evolution of Apioideae, even in the basal lineage Chamaesieae
(C.mallaeanum) (Guo etal., 2018) and the most spiny fruit lin-
eage Daucinae. As referred to in previous studies, D.decipiens
and D.edulis are endemic to Madeira, where has never been
connected to the mainland and lacks native large mammals that
could have acted as dispersing agents for epizoochorous spe-
cies, with the appendages of secondary ribs winged (Banasiak
etal., 2016; Wojewódzka etal., 2019). This indicated that
anemochory could be the best dispersal mode for species of
Apioideae if animal vectors are absent, and the formation of
wings for fruit ribs becomes an advantageous adaptation.
For most taxa of Daucinae and Torilidinae, epizoochory
might be the best dispersal mode. The dispersal unit of these
groups is not only individual mericarps but also the entire plant
(for Daucus), which combined with slightly or strongly com-
pressed endosperms of these fruits suggests an adaptation for
anemochory. Such dispersal strategies may be used to explain
the diversity and wide distribution of Daucus and the great dif-
ferentiation seen within D.carota. Species with less developed
ribs in Apioideae always have smaller fruits with a dense endo-
sperm (Bupleureae, Chamaesieae and Pimpinelleae), which
may be conducive to dispersal via gravity, and the smaller
size and mass of these fruits also appear to be an adaptation to
anemochory.
We mapped the fruit morphology and their dispersal modes
into our species tree and speculated that the fruits of Apioideae
might have undergone adaptive evolution and evolved in two
directions. One is towards forming fruits such as those of
Group C, which favour dispersal by water. The others tend to
develop winged fruits, like fruits of Group Aand GroupB,
adapted to dispersal by wind. The spiny fruits in Group B
with spines evolved from wings developing on secondary ribs
(Wojewódzka etal., 2019), suggesting epizoochory as a de-
rived mode of anemochory. Interestingly, the fruit characters of
Group Aand the Acronema Clade are distinctly homoplastic,
which may reect convergent adaptive evolution among these
lineages. However, when some exceptional cases, such as taxa
of Pimpinelleae, are considered, the entire plant habit must also
be taken into account. Normalized parameters are expected to
be proposed for the dispersal of Apioideae in the future, which
could integrate most of the characters that impact fruit dispersal.
These characters, as we suggest, should include fruit size, de-
velopment of ribs, habitat, type and population of local vectors,
and so on. All of these together determine the efciency and
modes of fruit dispersal.
An updated evolutionary divergence time scale of Apioideae
related to uplift of theQTP
Our divergence time estimate is credible because it depends
on a reliable phylogeny and displays short 95% HPDs. The 95%
HPDs on ages for our selected nodes are less than 13Myr except
the time of origin of Chamaesieae and Apioideae. Nevertheless,
our 95% HPD for the time of origin of Chamaesieae is
15.98Myr (42.72–58.7Mya), about half that inferred from
rps16 (30.8Myr) (Table3) (Calviño etal., 2016) and shares
part of the time scale in that study (37.2–68Mya). Additionally,
long 95% HPDs (some more than 30Myr) are given for the
six selected nodes of Apioideae lineages estimated by Calviño
etal. (2016), which were much longer than our results. Even
though 95% HPDs estimated by ITS were almost all shorter
than those from our study, the uncertain phylogeny restricted
their accuracy. We speculated that the evolution of Apioideae
might be related to uplift of the Qinghai–Tibetan Plateau (QTP)
based on this credible estimation of divergence time. Taxa of
Apioideae are distributed widely in the North Temperate zone,
and in East Asia, China is recognized as one of the most im-
portant distribution centres. The main centre of diversity of
Apioideae is south-west China, including Sichuan, Yunnan
and adjacent parts of Xizang A.R (Pimenov, 2017), across the
QTP and the Hengduan Mountains. We adopted the geological
scenario related to uplift of the QTP from Favre etal. (2015)
(Fig.5). It is undoubtedly very attractive to explore the complex
evolution of Apioideae, a group that evolved over almost the
entire period of the orogeny, probably closely related to these
geological events. Our estimated divergence time of Apioideae
(Fig.5) suggested that Apioideae originated at c. 56.64Mya
(95% HPD, 45.18–73.53Mya) slightly earlier than the colli-
sion of the Indian–Eurasian plates (van Hinsbergen etal., 2012;
Favre etal., 2015). The divergence time of two readily identi-
able earliest diverging lineages (Chamaesieae and Bupleureae)
of Apioideae in Asia probably indicates that Indian–Eurasian
collision did not cause a sharp differentiation of Apioideae taxa.
An alpine lineage (Pleurospermeae) originated at c. 41.15Mya
(95% HPD, 36.65–47.8Mya), consistent with the high eleva-
tion for the QTP during the early period of uplift (Favre etal.,
2015; Renner, 2016). The formation time of the morphologic-
ally rich taxa of distal branch D and the Physospermopsis Clade
was c. 33.07Mya (95% HPD, 27.44–40.01Mya), suggesting
that the divergence of their MRCA was occurred duringtime
Downloaded from https://academic.oup.com/aob/article-abstract/125/6/937/5722139 by Sichuan University user on 14 May 2020
Wen etal. — Transcriptome-based study of the phylogeny and evolution of Apioideae
950
of uplift of the QTP. The divergence time of the three succes-
sive sister groups (A, B, C) within branch D with deviations
of only c. 1.88Myr indicates that the ancestors of these spe-
cies might have undergone rapid evolutionary divergence, con-
sistent with our species tree with short branch length and low
quartet support (q1=0.49, q2=0.22, q3=0.3). The major
clades of these three groups (A, B, C) generally originated at
11.3–21.38Mya, possibly related not only to uplift but also to
climate changes (Favre etal., 2015). The evolution of some
Apioideae genera might have been driven by the nal period
of uplift (Favre etal., 2015), as inferred from the divergence
times (1.3–7.44Mya) within genera (Angelica, Pternopetalum
and Oenanthe), mainly during the orogeny of the Hengduan
Mountains and Qaidam basin. The origin of the two most con-
troversial lineages in our studies, Apieae (c. 12.29 Mya) and
the Sinodielsia Clade (c. 13.21 Mya), is relatively short (c. 0.92
Myr), consistent with rapid evolutionary change inferred from
our species tree. The Acronema Clade and Physospermopsis
Clade, which are almost exclusively found in east Asian (Zhou
etal., 2008; Zhou etal., 2009) and vary widely in morphological
characters, were suggested to have originated and diverged in
close association with the uplift of the QTP (Fig.5) and might
have undergone rapid evolutionary divergence. Overall, our
studies indicate that the QTP was likely to be an origin and dif-
ferentiation centre of Apioideae, and the taxa of these two east
Asian endemicclades (Acronema Clade and Physospermopsis
Clade) might provide details regarding the signicance of the
QTP orogeny on the evolution of Apioideae. The signicant
convergence of previously considered important morphological
characters indicated that taxa of Apioideae living in these re-
gions might be likely to undergo rapid radiations. Climate
models modied by orogeny are crucial to the biodiversity of
these regions, and the time frame of the orogeny provides clues
to the evolution of Apioideae. Fruit and dating analyses might
reect a convergent evolution between the Acronema Clade and
Group A.Unfortunately, we did not study the biogeography of
Apioideae in depth due to the small number of groups selected.
This is expected to be completed in future. Combining diver-
gence time and the relevant geological events, we infer that the
orogeny of the QTP might have triggered the diversication of
Apioideae, most taxa of which were still on the path of evolu-
tion, helping us to understand the complexity of the internal
classication of Apioideae.
CONCLUSION
Phylogenetic relationships among major lineages of Apioideae
have been a contentious issue due to the different molecular
markers used. Our study based on 3351 SCGs yielded a reli-
able species tree with ve branches with low quartet support
values, which suggested that rapid evolutionary divergence and
ILS may have been the main cause of conicts observed among
gene trees. Our species tree differs from the phylogenies in-
ferred from ITS and cpDNA (Zhou etal., 2009; Banasiak etal.,
2013; Calviño etal., 2016), with rearrangements among some
major lineages. Both our phylogeny and fruit analyses strongly
support that Scandiceae should be divided into at least two
major clades. One includes Daucinae and Torilidinae, which
should be restored as tribes according to both molecular and
morphological evidence. The presence of fruit wings seems
to be an optimization of Apioideae fruits over the present
phylogenies, which gave the group an adaptive advantage that
allowed it to diversify rapidly. The small number of groups may
cause some deviation in the divergence time estimates, although
it is credible to some degree in the context of a reliable phyl-
ogeny and short 95% HPDs. Molecular dating analysis based
on 11 concatenated SCGs suggested that Apioideae originated
around 56.64 Mya (95% HPD, 45.18–73.53 Mya), and the
QTP region acted as a probable diversity centre of Apioideae.
Uplift of the QTP and climatic changes probably drove rapid
speciation and diversication of Apioideae in the QTP region.
Our study shows that combining SCGs and coalescent-based
species tree estimation methods is a powerful approach that
provides more rened phylogenetic estimates for Apioideae
that were controversial based on small datasets.
SUPPLEMENTARYDATA
Supplementary data are available online at https://academic.
oup.com/aob and consist of the following.
Fig. S1: Quartet support for each branch in the species tree
of Apioideae.
Fig. S2: Species tree of Group Aestimated based on 3793
gene trees using ASTRAL-II.
Fig. S3: Quartet support for each branches in the species tree
of GroupA.
Fig. S4: Species tree of Group B estimated based on 4257
gene trees using ASTRAL-II.
Fig. S5: Quartet support for each branch in the species tree
of GroupB.
Fig. S6: Acluster of all the 3351 gene trees upon neighbour-
joining analysis.
Fig. S7: Species tree of the Bt branch generated using
ASTRAL-II based on 1184 SCGtrees.
Fig. S8: Species tree of the Bt branch generated using
MP-EST based on 1184 SCGtrees.
Fig. S9: Quartet support for each branch in the species tree
of the Bt branch.
Fig. S10: Species tree of the Ot branch generated using
ASTRAL-II based on 1260 SCGtrees.
Fig. S11: Quartet support for each branch in the species tree
of the Ot branch.
Fig. S12: Species tree of the Rt branch generated using
ASTRAL-II based on 405 SCGtrees.
Fig. S13: Quartet support for each branch in the species tree
of the Rt branch.
Fig. S14: Species tree of the Pt branch generated using
ASTRAL-II based on 502 SCGtrees.
Fig. S15: Species tree of the Pt branch generated using
MP-EST based on 502 SCGtrees.
Fig. S16: Quartet support for each branch in the species tree
of the Pt branch.
Fig. S17: Reconstruction of the two selected fruit characters
on the nal speciestree.
Table S1: Collection records of the 27 transcriptome datasets
sample.
Table S2: Final quartet score and normalized quartet score
for the datasets we explored.
Downloaded from https://academic.oup.com/aob/article-abstract/125/6/937/5722139 by Sichuan University user on 14 May 2020
Wen etal. — Transcriptome-based study of the phylogeny and evolution of Apioideae 951
FUNDING
This work was supported by the National Natural Science
Foundation of China (Grant Nos. 31872647), the fourth na-
tional survey of traditional Chinese medicine resources
(Grant No. 2019PC002), and the National Infrastructure of
Natural Resources for Science and Technology (Grant No.
2005DKA21403-JK).
ACKNOWLEDGEMENTS
We are grateful to Juan Li, Jiao Huang, Haoyu Hu and Chuan
Xie for help in collection of plant material; Yiqi Deng for as-
sembly of transcriptome datasets; and Junpei Chen for pro-
viding one fruit pattern. We thank Huaxi campus medicinal
botanical garden of Sichuan University and The Institute of
Medicinal Plant Development for collection of plant material.
We thank Novogene company for sequencing.
LITERATURECITED
AjaniY, AhmadA, CordesJM, WatsonMF, DownieSR. 2008. Phylogenetic
analysis of nrDNA ITS sequences reveals relationships within ve groups
of Iranian Apiaceae subfamily Apioideae. Taxon 57: 383–401.
AllioR, ScornavaccaC, NabholzB, ClamensAL, SperlingFAH, CondamineF.
2020. Whole genome shotgun phylogenomics resolves the pattern and timing
of swallowtail buttery evolution. Systematic Biology 69: 38–60.
AltınorduF, PeruzziL, YuY, HeXJ. 2016. A tool for the analysis of chromo-
somes: KaryoType. Taxon 65: 586–592.
BanasiakL, PiwczynskiM, UlinskiT, etal. 2013. Dispersal patterns in
space and time: a case study of Apiaceae subfamily Apioideae. Journal of
Biogeography 40: 1324–1335.
BanasiakL, WojewódzkaA, BaczyńskiJ, etal. 2016. Phylogeny of Apiaceae
subtribe Daucinae and the taxonomic delineation of its genera. Taxon 65:
563–585.
BellCD, SoltisDE, SoltisPS. 2010. The age and diversication of the angio-
sperms re-revisited. American Journal of Botany 97: 1296–1303.
BenthamG. 1867. Umbelliferae. In: BenthamandlG, HookerD, eds. Genera
Plantarum, vol. 1: 859–931.
BremerK, FriisEM, BremerB. 2004. Molecular phylogenetic dating
of Asterid owering plants shows Early Cretaceous diversication.
Systematic Biology 53: 496–505.
CalviñoCI, TilneyPM, vanWykBE, DownieSR. 2006. A molecular phylo-
genetic study of southern African Apiaceae. American Journal of Botany
93: 1828–1847.
CalviñoCI, MartínezSG, DownieSR. 2008. Morphology and biogeography
of Apiaceae subfamily Saniculoideae as inferred by phylogenetic analysis
of molecular data. American Journal of Botany 95: 196–214.
CalviñoCI, TeruelFE, DownieSR. 2016. The role of the southern hemi-
sphere in the evolutionary history of Apiaceae, a mostly north temperate
plant family. Journal of Biogeography 43: 398–409.
Capella-GutierrezS, Silla-MartinezJM, GabaldonT. 2009. trimAl: a tool
for automated alignment trimming in large-scale phylogenetic analyses.
Bioinformatics 25: 1972–1973.
Cerceau-LarrivalMTh. 1962. Plantules et pollens d’Ombelliferes. Memories
du Museum d’Histoire Naturelle, Serie B, Botanique 14: 1–166.
ChandlerGT, PlunkettGM. 2004. Evolution in Apiales: nuclear and chloro-
plast markers together in (almost) perfect harmony. Botanical Journal of
the Linnean Society 144: 123–147.
CloutierA, ScaktonTB, GraysonP, ClampM, BakerAJ, EdwardsS.
2019. Whole-genome analyses resolve the phylogeny of ightless birds
(Palaeognathae) in the presence of an empirical anomaly zone. Systematic
Biology 68: 937–955.
DegnanJH, RosenbergNA. 2006. Discordance of species trees with their
most likely gene trees. PLoS Genetics 2: e68.
DegnanJH, RosenbergNA. 2009. Gene tree discordance, phylogenetic in-
ference and the multispecies coalescent. Trends in Ecology and Evolution
24:332–340.
deQueirozA, GatesyJ. 2007. The supermatrix approach to systematics.
Trends in Ecology and Evolution 22: 34–41.
DeSmetR, AdamsKL, VandepoeleK, VanMontaguMCE, MaereS,
dePeerYV. 2013. Convergent gene loss following gene and genome du-
plications creates single-copy families in owering plants. Proceedings of
the National Academy of Sciences of the United States of America 110:
2898–903.
DownieSR, Katz-DownieDS. 1996a. A molecular phylogeny of Apiaceae
subfamily Apioideae: evidence from nuclear ribosomal DNA internal tran-
scribed spacer sequences. American Journal of Botany 83: 234–251.
DownieSR, Katz-DownieDS, ChoKJ. 1996b. Phylogenetic analysis of
Apiaceae subfamily Apioideae using nucleotide sequences from the
chloroplast rpoC1 intron. Molecular Phylogenetics and Evolution 6: 1–18.
DownieSR, RamanathS, Katz-DownieDS, LlanasE. 1998. Molecular sys-
tematics of Apiaceae subfamily Apioideae: phylogenetic analyses of nu-
clear ribosomal DNA internal transcribed spacer and plastid rpoC1 intron
sequences. American Journal of Botany 85: 563–591.
DownieSR, Katz-DownieDS, WatsonMF. 2000. A phylogeny of the
owering plant family Apiaceae based on chloroplast DNA rpl16 and
rpoC1 intron sequences: towards a suprageneric classication of sub-
family Apioideae. American Journal of Botany 87: 273–292.
DownieSR, PlunkettGM, WatsonMF, etal. 2001. Tribes and clades within
Apiaceae subfamily Apioideae: the contribution of molecular data.
Edinburgh Journal of Botany 58: 301–330.
DownieSR, HartmanRL, SunF-J, Katz-DownieDS. 2002. Polyphyly of
the spring-parsleys (Cymopterus): molecular and morphological evidence
suggests complex relationships among the perennial endemic genera
of western North American Apiaceae. Canadian Journal of Botany 80:
1295–1324.
DownieSR, SpalikK, Katz-DownieDS, ReduronJP. 2010. Major clades
within Apiaceae subfamily Apioideae as inferred by phylogenetic analysis
of nrDNA ITS sequences. Plant Diversity and Evolution 128: 111–136.
DrudeCGO. 1898. Umbelliferae. In: EnglerA, PrantlK, eds. Die Natürlichen
Panzen-familien 3. Leipzig: W. Engelmann, 63–250.
DrummondAJ, SuchardMA, XieD, RambautA. 2012. Bayesian
phylogenetics with BEAUti and the BEAST 1.7. Molecular Biology and
Evolution 29: 1969–1973.
DuarteJM, WallPK, EdgerPP, etal. 2010. Identication of shared single
copy nuclear genes in Arabidopsis, Populus, Vitis and Oryza and their
phylogenetic utility across various taxonomic levels. BMC Evolutionary
Biology 10: 61.
DunnCW, HejnolA, MatusDQ, etal. 2008. Broad phylogenomic sampling
improves resolution of the animal tree of life. Nature 452: 745–749.
EdwardsSV, LiuL, PearlDK. 2007. High-resolution species trees without
concatenation. Proceedings of the National Academy of Sciences of the
United States of America 104: 5936–5941.
FavreA, PaeckertM, PaulsSU, etal. 2015. The role of the uplift of the
Qinghai-Tibetan Plateau for the evolution of Tibetan biotas. Biological
Reviews 90: 236–253.
FernándezM, EzcurraC, CalviñoCI. 2017a. Chloroplast and ITS
phylogenies to understand the evolutionary history of southern South
American Azorella, Laretia and Mulinum (Azorelloideae, Apiaceae).
Molecular Phylogenetics and Evolution 108: 1–21.
FernándezM, EzcurraC, CalviñoCI. 2017b. Species limits and mor-
phometric and environmental variation within the South Andean and
Patagonian Mulinum spinosum species-group (Apiaceae-Azorelloideae).
Systematics and Biodiversity 15: 489–505.
FischerS, BrunkBP, ChenF, etal. 2011. Using OrthoMCL to assign pro-
teins to OrthoMCL‐DB groups or to cluster proteomes into new ortholog
groups. Current Protocols in Bioinformatics 35: 6.12.1–6.12.19.
FlouriT, JiaoXY, RannalaB, YangZH. 2018. Species tree inference with
BPP using genomic sequences and the multispecies coalescent. Molecular
Biology and Evolution 35: 2585–2593.
FuentesS, Vivian-SmithA. 2009. Fertilisation and fruit initiation. In:
ØstergaardL, ed. Annual Plant Reviews Volume 38: Fruit development
and seed dispersal. Oxford: John Wiley and Sons, 107–171.
Gruas-CavagnettoC, Cerceau-LarrivalM-T. 1984. Apports des pol-
lens fossiles d’Ombellif eres a la connaissance paléoécologique et
paléoclimatique de l’Eocene français. Review of Palaeobotany and
Palynology 40: 317–345.
GuoXL, WangCB, WenJ, ZhouSD, HeXJ. 2018. Phylogeny of Chinese
Chamaesium (Apiaceae: Apioideae) inferred from ITS, cpDNA and mor-
phological characters. Phytotaxa 376: 1–016.
Downloaded from https://academic.oup.com/aob/article-abstract/125/6/937/5722139 by Sichuan University user on 14 May 2020
Wen etal. — Transcriptome-based study of the phylogeny and evolution of Apioideae
952
HanF, PengY, XuL, XiaoP. 2014. Identication, characterization, and util-
ization of single copy genes in 29 angiosperm genomes. BMC Genomics
15: 504.
HeledJ, DrummondAJ. 2010. Bayesian inference of species trees from
multilocus data. Molecular Biology and Evolution 27: 570–580.
HeywoodVH. 1986. The Umbelliferae – an impossible family? Symbolae
Botanicae Upsalienses 26: 73–80.
HochbachA, SchneiderJ, RöserM. 2015. A multi-locus analysis of phylo-
genetic relationships within grass subfamily Pooideae (Poaceae) inferred
from sequences of nuclear single copy gene regions compared with plastid
DNA. Molecular Phylogenetics and Evolution 87: 14–27.
HohnaS, LandisMJ, HeathTA, etal. 2016. RevBayes: Bayesian phylo-
genetic inference using graphical models and an interactive model-
specication language. Systematic Biology 65: 726–736.
IorizzoM, EllisonS, SenalikD, etal. 2016. A high-quality carrot genome
assembly provides new insights into carotenoid accumulation and asterid
genome evolution. Nature Genetics 48: 657–666.
JianSG, SoltisPS, GitzendannerMA, etal. 2008. Resolving an ancient,
rapid radiation in Saxifragales. Systematic Biology 57: 38–57.
JongejansE, TeleniusA. 2001. Field experiments on seed dispersal by wind in
ten umbelliferous species (Apiaceae). Plant Ecology 152: 67–78.
Katz-DownieDS, Valiejo-RomanCM, TerentievaEI, etal. 1999. Towards a
molecular phylogeny of Apiaceae subfamily Apioideae: additional infor-
mation from nuclear ribosomal DNA ITS sequences. Plant Systematics
and Evolution 216: 167–195.
KochWDJ. 1824. Generum tribuumque plantarum Umbelliferarum nova
dispositio. Nova Acta Academiae Caesareae Leopoldino Carolinae
Germanicae Naturae Curiosorum 12: 55–156.
Kozo-PoljanskyBM. 1916. Sciadopnytorum systematis lineamenta. Bulletin
of Moscow Society of Naturalist 29: 93–222.
KubatkoLS, DegnanJH. 2007. Inconsistency of phylogenetic estimates from
concatenated data under coalescence. Systematic Biology 56: 17–24.
LeeBY, DownieSR. 2000. Phylogenetic analysis of cpDNA restriction
sites and rps16 intron sequences reveals relationships among Apiaceae
tribes Caucalideae, Scandiceae and related taxa. Plant Systematics and
Evolution 221: 35–60.
LeeBY, LevinGA, DownieSR. 2001. Relationships within the spiny-fruited
umbellifers (Scandiceae subtribes Daucinae and Torilidinae) as assessed
by phylogenetic analysis of morphological characters. Systematic Botany
26: 622–642.
Leebens-MackJH, BarkerMS, CarpenterEJ, etal. 2019. One thousand
plant transcriptomes and the phylogenomics of green plants. Nature 574:
679–685.
LiM, WunderJ, BissoliG, etal. 2008. Development of COS genes as uni-
versally ampliable markers for phylogenetic reconstructions of closely
related plant species. Cladistics 24: 727–745.
LinnaeusC. 1753. Species Plantarum. Stockholm: Laurentii Salvii.
LiuL, YuLL, PearlDK, EdwardsSV. 2009. Estimating species phylogenies
using coalescence times among sequences. Systematic Biology 58:
468–477.
LiuL, YuLL, EdwardsSV. 2010. A maximum pseudo-likelihood approach for
estimating species trees under the coalescent model. BMC Evolutionary
Biology 10: 302.
LiuL, YuL. 2011. Estimating species trees from unrooted gene trees.
Systematic Biology 60: 661–667.
LiuM, WykBV, TilneyPM, PlunkettGM, LowryPP, MageeAR. 2016.
The phylogenetic signicance of fruit structural variation in the tribe
Heteromorpheae (Apiaceae). Pakistan Journal of Botany 48: 201–210.
LiuMR, DownieSR. 2017. The phylogenetic signicance of fruit anatomical
and micromorphological structures in Chinese Heracleum species and re-
lated taxa (Apiaceae). Systematic Botany 42: 313–325.
LyskovDF, DegtjarevaGV, SamigullinTH, PimenovMG. 2017. The revi-
sion of Prangos, subsections Koelzella and Fedtschenkoana, (Apiaceae)
with some notes to phylogeny and biogeography of the genus: molecular
and morphological evidences. Plant Systematics and Evolution 313:
815–826.
MaddisonWP. 1997. Gene trees in species trees. Systematic Biology 46:
523–536.
MaddisonWP, MaddisonDR. 2018. Mesquite: a modular system for evolu-
tionary analysis. Version 3.51. Available at: http://www.mesquiteproject.org
MaoKS, RuhsamM, MaYZ, etal. 2019. A transcriptome-based resolution
for a key taxonomic controversy in Cupressaceae. Annals of Botany 123:
221.
MirarabS, ReazR, BayzidMS, ZimmermannT, SwensonMS, WarnowT.
2014. ASTRAL: genome-scale coalescent-based species tree estimation.
Bioinformatics 30: i541–i548.
MirarabS, WarnowT. 2015. ASTRAL-II: coalescent-based species tree esti-
mation with many hundreds of taxa and thousands of genes. Bioinformatics
32: 44–52.
MorisonR. 1672. Plantarum umbelliferarum distributio nova. Oxford.
MosselE, VigodaE. 2005. Phylogenetic MCMC algorithms are misleading on
mixtures of trees. Science 309: 2207–2209.
NicolasAN, PlunkettGM. 2014. Diversication times and biogeographic pat-
terns in Apiales. The Botanical Review 80: 30–58.
Jiménez-MejíasP, VargasP. 2015. Taxonomy of the tribe Apieae (Apiaceae)
revisited as revealed by molecular phylogenies and morphological charac-
ters. Phytotaxa 212: 57–79.
PimenovMG, LeonovMV. 1993. The genera of the Umbelliferae.
ANomenclator. Kew: Royal Botanical Gardens.
PimenovMG. 2017. Updated checklist of Chinese Umbelliferae: nomencla-
ture, synonymy, typication, distribution. Turczaninowia 20: 106–239.
PlunkettGM, SoltisDE, SoltisPS. 1996a. Higher level relationships of
Apiales (Apiaceae and Araliaceae) based on rbcL sequences. American
Journal of Botany 83: 499–515.
PlunkettGM, SoltisDE, SoltisPS. 1996b. Evolutionary patterns in Apiaceae:
inferences based on matK sequences data. Systematic Botany 21: 477–495.
PlunkettGM, SoltisDE, SoltisPS. 1997. Clarication of the relationship be-
tween Apiaceae and Araliaceae based on matK and rbcL sequence data.
American Journal of Botany 84: 565–580.
PlunkettGM, DownieSR. 1999. Major lineages within Apiaceae subfamily
Apioideae: a comparison of chloroplast restriction site and DNA sequence
data. American Journal of Botany 86: 1014–1026.
PlunkettGM, ChandlerGT, Lowry II PP, PinneySM, SprenkleTS. 2004.
Recent advances in understanding Apiales and a revised classication.
South African Journal of Botany 70: 371–381.
PosadaD, CrandallKA. 1998. MODELTEST: testing the model of DNA sub-
stitution. Bioinformatics 14: 817–818.
RannalaB, YangZH. 2017. Efcient Bayesian species tree inference under
the multispecies coalescent. Systematic Biology 66: 823–842.
RennerSS. 2016. Available data point to a 4-km-high Tibetan Plateau by 40
Ma, but 100 molecular-clock papers have linked supposed recent uplift to
young node ages. Journal of Biogeography 43: 1479–1487.
RobinsonDF, FouldsLR. 1981. Comparison of phylogenetic trees.
Mathematical Biosciences 53: 131–147.
RochS, SteelM. 2014. Likelihood-based tree reconstruction on a concat-
enation of aligned sequence data sets can be statistically inconsistent.
Theoretical Population Biology 100: 56–62.
RokasA, WilliamsBL, KingN, CarrollSB. 2003. Genome-scale approaches
to resolving incongruence in molecular phylogenies. Nature 425: 798–804.
RömermannC, TackenbergO, PoschlodP. 2005. How to predict attachment
potential of seeds to sheep and cattle coat from simple morphological seed
traits. Oikos 110: 219–230.
RonquistF, TeslenkoM, vanderMarkP, etal. 2012. MrBayes 3.2: efcient
Bayesian phylogenetic inference and model choice across a large model
space. Systematic Biology 61: 539–542.
SayyariE, MirarabS. 2016. Fast coalescent-based computation of local
branch support from quartet frequencies. Molecular Biology and Evolution
33: 1654–1668.
ScallyA, DutheilJY, HillierLDW, etal. 2012. Insights into hominid evolu-
tion from the gorilla genome sequence. Nature 483: 169–175.
ShawTI, RuanZ, GlennTC, LiuL. 2013. STRAW: Species TRee Analysis
Web server. Nucleic Acids Research 41: W238–W241.
ShneyerVS, BorschtschfnkoGP, PimenovMG, LeonovMV. 1992. The
tribe Smyrnieae (Umbelliferae) in the light of serotaxonomical analysis.
Plant Systematics and Evolution 182: 135–148.
SmallRL, CronnRC, WendelJF. 2004: Use of nuclear genes for phylogeny
reconstruction in plants. Australian Systematic Botany 17: 145–170.
SongS, LiuL, EdwardsSV, WuS. 2012. Resolving conict in eutherian
mammal phylogeny using phylogenomics and the multispecies coalescent
model. Proceedings of the National Academy of Sciences of the United
States of America 109: 14942–14947.
SpalikK, DownieSR. 2001. The utility of morphological characters for
inferring phylogeny in Scandiceae subtribe Scandicinae (Apiaceae).
Annals of the Missouri Botanical Garden 88: 270–301.
SpalikK, ReduronJP, DownieSR. 2004. The phylogenetic position of
Peucedanum sensu lato and allied genera and their placement in tribe
Downloaded from https://academic.oup.com/aob/article-abstract/125/6/937/5722139 by Sichuan University user on 14 May 2020
Wen etal. — Transcriptome-based study of the phylogeny and evolution of Apioideae 953
Selineae (Apiaceae, subfamily Apioideae). Plant Systematics and
Evolution 243: 189–210.
SpalikK, DownieSR. 2007. Intercontinental disjunctions in Cryptotaenia
(Apiaceae, Oenantheae): an appraisal using molecular data. Journal of
Biogeography 34: 2039–2054.
SpalikK, PiwczyńskiM, DandersonCA, Kurzyna-MłynikR, BoneTS,
DownieSR. 2010. Amphitropic amphiantarctic disjunctions in Apiaceae
subfamily Apioideae. Journal of Biogeography 37: 1977–1994.
StamatakisA. 2014. RAxML version 8: a tool for phylogenetic analysis and
post-analysis of large phylogenies. Bioinformatics 30: 1312–1313.
TackenbergO, RömermannC, ThompsonK, PoschlodP. 2006. What does
diaspore morphology tell us about external animal dispersal? Evidence
from standardized experiments measuring seed retention on animal-coats.
Basic and Applied Ecology 7: 45–58.
TeasdaleLC, KöhlerF, MurrayKD, O’HaraT, MoussalliA .
2016. Identication and qualication of 500 nuclear, single-copy,
orthologous genes for the Eupulmonata (Gastropoda) using transcrip-
tome sequencing and exon-capture. Molecular Ecology Resources 16:
1107–1123.
TheobaldWL. 1971. Comparative anatomical and developmental studies in
the Umbelliferae. In: HeywoodVH, ed. The biology and chemistry of the
Umbelliferae. London: Academic Press, 177–197.
vanHinsbergenDJJ, LippertPC, Dupont-NivetG, etal. 2012. Greater
India Basin hypothesis and a two-stage Cenozoic collision between India
and Asia. Proceedings of the National Academy of Sciences of the United
States of America 109: 7659–7664.
WeitzelC, RønstedN, SpalikK, SimonsenHT. 2014. Resurrecting deadly
carrots: towards a revision of Thapsia (Apiaceae) based on phylogenetic
analysis of nrITS sequences and chemical proles. Botanical Journal of
the Linnean Society 174: 620–636.
WilliamsCF. 1994. Genetic consequences of seed dispersal in three sym-
patric forest herbs. II. Microspatial genetic structure within populations.
Evolution 48: 1959–1972.
WilliamsCF, GuriesRP. 1994. Genetic consequences of seed dispersal in
three sympatric forest herbs. I.Hierarchical population-genetic structure.
Evolution 48: 791–805.
WilliamJ, BallardO. 1996. Combining data in phylogenetic analysis. Trends
in Ecology and Evolution 11: 334.
WojewódzkaA, BaczyńskiJ, BanasiakŁ, etal. 2019. Evolutionary shifts in
fruit dispersal syndromes in Apiaceae tribe Scandiceae. Plant Systematics
and Evolution 305: 401–414.
WuF, MuellerLA, CrouzillatD, PétiardV, TanksleySD. 2006. Combining
bioinformatics and phylogenetics to identify large sets of single-copy
orthologous genes (COSII) for comparative, evolutionary and systematic
studies: a test case in the euasterid plant clade. Genetics 174: 1407–1420.
XiangYZ, HuangCH, HuY, etal. 2017. Evolution of Rosaceae fruit types
based on nuclear phylogeny in the context of geological times and genome
duplication. Molecular Biology and Evolution 34: 262–281.
YangZ, RannalaB. 2014. Unguided species delimitation using DNA sequence
data from multiple loci. Molecular Biology and Evolution 31: 3125–3135.
YuY, BlairC, HeXJ. 2020. RASP 4: ancestral state reconstruction tool for
multiple genes and characters. Molecular Biology and Evolution 37:
604–606.
YuanYW, LiuC, MarxHE, OlmsteadRG. 2009.The pentatrico peptide
repeat (PPR) gene family, a tremendous resource for plant phylogenetic
studies. New Phytologist 182: 272–283.
ZakharovaEA, KljuykovEV, DegtjarevaGV, SamigullinTH,
UkrainskayaUA, DownieSR. 2016. A taxonomic study of the genus
Hellenocarum H.wolff (Umbelliferae-Apioideae) based on morphology,
fruit anatomy, and molecular data. Turkish Journal of Botany 40: 176–193.
ZengLP, ZhangQ, SunRR, KongHZ, ZhangN, MaH. 2014. Resolution of
deep angiosperm phylogeny using conserved nuclear genes and estimates
of early divergence times. Nature Communications 5: 4956.
ZhongB, LiuL, YanZ, PennyD. 2013. Origin of land plants using the
multispecies coalescent model. Trends in Plant Science 18: 492–495.
ZhouJ, PengH, DownieSR, LiuZW, GongX. 2008. A molecular phylogeny
of Chinese Apiaceae subfamily Apioideae inferred from nuclear ribosomal
DNA internal transcribed spacer sequences. Taxon 57: 402–416.
ZhouJ, GongX, DownieSR, PengH. 2009. Towards a more robust molecular
phylogeny of Chinese Apiaceae subfamily Apioideae: additional evi-
dence from nrDNA ITS and cpDNA intron (rpl16 and rps16) sequences.
Molecular Phylogenetics and Evolution 53: 56–68.
Downloaded from https://academic.oup.com/aob/article-abstract/125/6/937/5722139 by Sichuan University user on 14 May 2020
