Plant Systematics and Evolution (2020) 306:9
Molecular systematics ofRosoideae (Rosaceae)
XunChen1,2 · JinluLi1,2· TaoCheng2,3· WenZhang2,3· YanleiLiu2,3· PingWu2,3· XueyingYang4· LingWang1·
Received: 14 August 2019 / Accepted: 8 January 2020
© Springer-Verlag GmbH Austria, part of Springer Nature 2020
The four-subfamily subdivision of Rosaceae has been recently replaced by a three-subfamily scheme. The re-circumscribed
Rosoideae lacks a solid and well-resolved phylogeny on which a classiﬁcation can be based. In this study, we sampled 56
genera presumably belonging to Rosoideae and 10 genera belonging to other subfamilies or families and used 12 chloroplast
regions (matK, rbcL, trnL, trnL–F, ndhF, ycf1, trnC–ycf6, trnS–G, trnS, psbA–trnH, rpoC1 and trnS–ycf9) to reconstruct
their phylogeny. Our results conﬁrmed (1) the exclusion of Rhodotypos and Kerria from Rosoideae and their inclusion in the
subfamily Amygdaloideae and (2) the exclusion of Chamaebatia, Cercocarpus, Dryas and Purshia (including Cowania) from
Rosoideae and their inclusion in Dryadoideae, the sister subfamily of Rosoideae. Within Rosoideae, there are six strongly
supported lineages that correspond to six tribes: Ulmarieae, Colurieae, Rubeae, Roseae, Agrimonieae and Potentilleae. We
dated the divergence of Rosoideae back to approximately 69.77 million years ago (Mya; 95% HPD = 61.28–78.33 Mya)
and that of the tribes within Rosoideae to from 10.42 to 40.02 million yearsago (Mya; 95% HPD = 4.73–59.08 Mya). The
subfamily is probably of North American and Asian origin and thrives in the northern hemisphere, especially in Asia. After
re-circumscriptions of several genera, there are 36 genera recognized in Rosoideae.
Keywords Biogeography· Divergence times· Phylogeny· Rosoideae
Rosaceae A.L.Jussieu comprises 2825–3500 species
(belonging to 95–125 genera) of trees, shrubs or herbs.
The group is cosmopolitan, mostly occurring in temperate
regions in the northern hemisphere. Many species in this
family are of great economic value, such as strawberries,
apples, cherries, pears and roses.
Several diﬀerent taxonomic treatments regarding the sub-
division of Rosaceae have been published. The most widely
adopted classiﬁcation ever (Schulze-Menz 1964) divides the
family into four subfamilies mostly based on fruit types:
Amygdaloideae (Prunoideae), Maloideae, Rosoideae and
Spiraeoideae. Molecular phylogenetic analyses based on
plastid DNA sequences (Morgan etal. 1994; Potter etal.
2002) demonstrated that Schulze-Menz’s (1964) Maloideae,
Prunoideae and most Rosoideae are monophyletic but that
Spiraeoideae is clearly polyphyletic. Based on phylogenetic
analyses of six nuclear and four chloroplast genes, Potter
etal. (2007) subdivided Rosaceae into three subfamilies:
Dryadoideae (comprising nitrogen-ﬁxing species formerly
classiﬁed in Rosoideae); Rosoideae and Amygdaloideae.
Handling Editor: Christian Parisod.
Electronic supplementary material The online version of this
article (https ://doi.org/10.1007/s0060 6-020-01629 -z) contains
supplementary material, which is available to authorized users.
* Xueying Yang
* Ling Wang
* Shiliang Zhou
1 College ofLandscape Architecture, Northeast Forestry
University, Harbin150040, China
2 State Key Laboratory ofSystematic andEvolutionary
Botany, Institute ofBotany, Chinese Academy ofSciences,
3 College ofLife Sciences, University ofChinese Academy
ofSciences, Beijing100049, China
4 National Engineering Laboratory forForensic Science,
Key Laboratory ofForensic Genetics, Institute ofForensic
Science, Ministry ofPublic Security, Beijing100038, China
X. Chen etal.
9 Page 2 of 12
Amygdaloideae includes former Spiraeoideae, Prunoideae,
Maloideae and a few genera formerly in Rosoideae.
The re-circumscribed Rosoideae contains approximately
56 genera and 1500 species of herbs, shrubs and rarely trees.
The fruit type is an achene (torus sometimes ﬂeshy and ber-
rylike) or a drupelet. Plants of the subfamily Rosoideae
occur all over the world, but mainly in the temperate North-
ern Hemisphere, forming intercontinental biogeographic
patterns of disjunction (Li 1952; Raven 1972; Thorne
1972). Among the intercontinental disjunctions, the eastern
Asian and the North American ﬂoristic disjunction received
much more attentions (Wen 1998, 1999; Xiang etal. 2000).
However, several genera in Rosoideae such as Acomastylis
Greene, Chamaerhodos Bge. and Waldsteinia Willd. are not
among the 65 genera of seed plants displaying such disjunc-
tion (Wen 1999; Xiang etal. 2000). The North Atlantic Land
Bridges (NALB), which connected eastern North America to
Europe via Greenland and the Bering Land Bridge (BLB),
which connected western North America and northeastern
Asia were used to explain the intercontinental disjunctions
(Tiﬀney 1985a, b; Tiﬀney and Manchester 2001; Wen 1999;
Wen etal. 2010). Divergence times of taxa are crucial to
understand which bridge was used by diﬀerent plant groups
because NALB was closed in the early Eocene (50 Mya) but
BLB remained open until ~ 3.5 Mya. The oldest fossil record
within Rosoideae is Rosa germerensis in the early Eocene
(cf. Palaeobiology database http://paleo db.org). However,
the subfamily was molecularly dated back to 50–80 Mya
(Chin etal. 2014; Xiang etal. 2017; Zhang etal. 2017).
Unfortunately, divergence time estimations of the major
clades in Rosoideae are still lacking. What seems more
important to ﬁll the blanks ﬁrst is to construct solid phylo-
genetic relationships within Rosoideae.
Rosoideae was divided into diﬀerent number of tribes
based on morphology (Eriksson etal. 2003; Hutchinson
1964; Kalkman 2004; Schulze-Menz 1964; Takhtajan
1997). The DNA sequence-based studies (Potter etal.
2007; Xiang etal. 2017) suggested six tribes in Rosoideae,
i.e., Agrimonieae, Colurieae, Potentilleae, Roseae,
Rubeae and Ulmarieae. The relationships among some
tribes are still unclear. For example, Roseae was shown
to be sistership to Agrimonieae + Potentilleae (Eriksson
etal. 2003; Xiang etal. 2017) or a relationship between
Roseae + Potentilleae and Agrimonieae (Potter etal. 2007;
Zhang etal. 2017). More phylogenetic and taxonomic
problems within Rosoideae are at generic level. Potentilla
L. is not monophyletic based on molecular phylogenies.
Eriksson etal. (1998) used ITS sequences to investigate
the monophyly of Potentilla in broad sense and found
that Potentilla is a paraphyletic or polyphyletic genus.
The addition of the chloroplast trnL–trnF signiﬁcantly
improved the resolution of molecular trees, and the genera
Horkelia Cham. & Schltdl. and Ivesia Torr. & A.Gray were
found to be nested within Potentilla (Eriksson etal. 2003).
Still others found that the relationships among Alchemilla
L., Aphanes L. and Lachemilla Focke were uncertain
(Eriksson etal. 2003; Gehrke etal. 2008). These phylo-
genetic uncertainties are probably due to both insuﬃcient
information of DNA sequences and inadequate sampling
of genera, and it is unlikely to provide a comprehensive
and solid phylogeny for the taxonomy of Rosoideae.
In this study, we aim to (1) reconstruct the phylogeny at
generic level within the subfamily Rosoideae; (2) estimate
the divergence times of major clades; and (3) build the
historical biogeography of the subfamily.
Plant sampling andgenomic DNA extraction
A total of 48 species from 27 genera in Rosaceae, includ-
ing 39 species from 19 genera in Rosoideae, 3 species
from 2 genera in Prunoideae (= Amygdaloideae), 3 species
from 3 genera in Spiraeoideae and 3 species from 3 genera
in Maloideae (Online Resource 1) were collected based
on the taxonomy of Rosaceae in Flora of China (Lu etal.
2003). Leaves of each species were dried with silica gel.
Genomic DNA was isolated from the dried leaves using
the mCTAB (modiﬁed cetyltrimethylammonium bromide)
protocol (Li etal. 2013).
Chloroplast DNA fragment amplication
Six chloroplast regions, trnL–trnF, rbcL, matK, ndhF, ycf1
and rpoC1, were ampliﬁed using primers (Online Resource
2) described previously (Dong etal. 2012, 2014, 2015;
Gardens 2007; Taberlet etal. 1991; Yu etal. 2011). PCR
was performed in 20-µL ampliﬁcation reactions containing
2 µL of DNA solution (20ng), 2 µL of 10× PCR buﬀer,
2 µL dNTP mix (0.2mM), 1 µL of each primer (5mM),
0.2 µL of Taq DNA polymerase (2.5 U/µL) and 11.8 µL of
ddH2O. The PCR thermal cycling proﬁle was performed
with 3min at 94°C followed by 35 cycles of 30s at 94°C,
40s at 52°C and 1min at 72°C, with a ﬁnal extension of
10min at 72°C. The PCR products were puriﬁed using
PEG8000 and sequenced on an ABI 3730xl DNA analyzer
(Majorbio Company, Beijing, China). The sequences were
manually assembled and edited in Sequencher, version 4.7
(Gene Codes Corporation, Ann Arbor, Michigan, USA).
Molecular systematics of Rosoideae (Rosaceae)
Page 3 of 12 9
The DNA sequences of each species in Rosoideae were
downloaded from GenBank (Online Resource 3). Misla-
beled sequences were identiﬁed using maximum parsimony
(MP) analysis method with PAUP* (Swoﬀord 2002) and
discarded. Finally, a total of 121 taxa, including 54 genera
of Rosoideae, were included in the phylogenetic analyses
in combination with newly collected sequence data. The
datasets were aligned using MAFFT (Katoh and Standley
2013) and manually adjusted using SE-Al 2.0 (Rambaut
2002). DnaSP 5.0 (Librado and Rozas 2009) was used to
assess the variability of markers, and PAUP 4.0b10 (Swof-
ford 2002) was used to judge the phylogenetic performance
of each gene partition (Table1). Ultimately, 12 chloroplast
regions of 149 species representing 56 presumable genera
in Rosoideae remained in the phylogenetic analysis. After
concatenating these datasets into one big data matrix, two
primary data matrices were constructed: (1) matrix A with
169 taxa for phylogenetic reconstruction and (2) matrix B
with 130 taxa for divergence time and phylogeographical
inference for reducing the computational burden. ModelTest,
version 3.7 (Posada and Crandall 1998), was used to deter-
mine the best ﬁtting model for the concatenated matrix under
the Akaike information criterion (AIC, Posada and Buckley
2004). Phylogenetic analyses of matrix A were performed
using MrBayes, version 3.2.2 (Ronquist and Huelsenbeck
2003), for Bayesian inferences (BI), PAUP* 4.0b10 (Swof-
ford 2002) for maximum parsimony (MP) analyses and
RAxML-HPC2, version 8.2.9 (Stamatakis 2014), for maxi-
mum likelihood analyses. Bayesian analyses were run for
10,000,000 generations with trees sampled every 1000 gen-
erations. The runs were stopped when the standard devia-
tion of split frequencies (SDSF) fell below 0.01. In the MP
analyses, all gaps were considered as missing data, and all
characters were unordered and equally weighted. For heu-
ristic searches, 1000 random sequence addition replicates
were conducted with tree bisection reconnection branch
(TBR) swapping. When running 1000 bootstrap support
(BS) replicates, the random addition sequence was limited
to 100. The ML analyses employed 1000 bootstrap itera-
tions, with default settings for the optimization of individual
per-site substitution rates. We performed all ﬁnal runs on the
CIPRES Science Gateway (https ://www.phylo .org/porta l2/,
Miller etal. 2010).
Divergence time estimation
To reduce the computational burden, the taxa with more
gene regions in the genera comprising more than three spe-
cies were selected. The topologies of this reduced matrix
B are similar to those of the larger matrix A. Although this
reduced matrix may introduce biases in terms of downstream
analysis (Hillis 1998), the general broad geographic patterns
and divergence times within the subfamily are likely kept
normal. The matrix B was analyzed with BEAST, version
1.8.3 (Drummond etal. 2012). An uncorrelated lognormal
relaxed clock model was used to estimate the rate changes
based on an initial likelihood ratio test. By restricting the
age of Rosales within [96, 90] Mya and calibrating the age
of Rosoideae with two other fossils (Table2), the diver-
gence times of Rosoideae were estimated together with out-
groups. The fossils used to calibrate the molecular clock
were selected from the online fossil resources (Paleobiol-
ogy database, <http://paleo biodb .org>) and literature survey
(Table2). The divergence times of Rosoideae were estimated
together with outgroups (Online Resource 4). BEAST run
was on the CIPRES Web server (http://www.phylo .org/).
Table 1 Variability of twelve
chloroplast regions in Rosoideae
and their maximum parsimony
N number of species; La aligned length; Sc sites considered; S number of polymorphic sites, excluding
sites with missing data; NH number of haplotypes; n nucleotide diversity per site after exclusion of sites
with missing data; k average number of nucleotide diﬀerences; Vs variable site; Is parsimony-informative
site; L the tree length; CI consistency index; and RI retention index
Chloroplast genes N La Sc S NH πk Vs Is L CI RI
matK 33 1306 820 232 28 0.06136 50.318 347 208 517 0.795 0.891
ndhF 23 1999 1090 194 21 0.04822 44.664 445 260 696 0.767 0.792
psbA–trnH 11 576 246 47 11 0.05255 12.927 95 25 116 0.914 0.818
rbcL 69 1376 460 77 44 0.03484 16.045 231 182 460 0.591 0.851
rpoCl 13 487 380 26 11 0.01849 7.026 36 20 44 0.818 0.800
trnC–ycf6 25 1100 674 145 25 0.05054 34.067 182 102 228 0.89 0.949
trnL 64 1423 445 197 51 0.08362 37.212 461 314 797 0.754 0.923
trnL–F 66 1332 399 171 48 0.09915 34.94 409 304 661 0.784 0.933
trnS 29 548 286 70 24 0.04801 13.732 101 45 125 0.872 0.942
trnS–G 20 1395 851 111 18 0.03595 30.595 189 133 223 0.897 0.953
trnS–ycf9 19 537 432 68 17 0.03698 15.977 76 37 95 0.863 0.900
ycf1 26 930 832 323 21 0.09753 81.148 345 242 614 0.748 0.838
X. Chen etal.
9 Page 4 of 12
Four independent runs of 300 million generations were
conducted by sampling one tree every 30,000 generations.
Other key settings were set as follows: tree prior = specia-
tion, Yule process; and substitution model = GTR + I+G.
Tracer, version 1.5.0(Rambaut and Drummond 2009), was
used to monitor the Bayesian search until ESS > 200. Log
Combiner (Rambaut and Drummond 2010a) and Tree Anno-
tator, version 1.5.4 (Rambaut and Drummond 2010b), were
used to calculate the maximum clade credibility tree.
Reconstruction ofancestral areas
The same reduced matrix B was used to reconstruct the
ancestral areas. To do this analysis, we collected the geo-
graphical distributions by genera by consulting to the infor-
mation collected from Encyclopedia of Life (EOL, http://
www.eol.org/) and Global Biodiversity Information Facil-
ity (GBIF, https ://www.gbif.org/). The geographical dis-
tributions we used in this study are summarized in Online
Resource 5. The geographical distribution of the extant spe-
cies in Rosoideae was divided into eight biogeographic areas:
(A) North America; (B) South America; (C) Africa; (D)
Asia; (E) Europe; (F) Mediterranean; (G) Oceania; and (H)
Arctic. A statistical dispersal–vicariance analysis (S-DIVA,
Yu etal. 2010) was performed to estimate the ancestral areas
of Rosoideae using RASP, version 3.1 (Yu etal. 2015), to
obtain its dispersal routes. The phylogenetic trees (contains
twenty thousand trees and the condensed tree) derived from
the BEAST analysis were used for S-DIVA. The analysis
included 56 genera in the Rosoideae clades and 10 outgroup
taxa with the number of max area of each node = 2.
Phylogenetic relationships inRosoideae
The ﬁnal aligned matrix A comprised 169 taxa and 26,474
characters after removing ambiguous aligned characters. The
features of each locus and their maximum parsimony tree
scores are shown in Table1. The ndhF region is the longest,
and the rpoC1 region is the shortest; ycf1 is the most vari-
able, and rpoC1 is the least variable.
The trees based on the cpDNA data built by MP, ML and
BI were nearly identical (Fig.1, Online Resources 6–8).
Three clades within Rosaceae (bp = 100, pp = 1, Online
Resource 6) corresponding to three subfamilies, Amygda-
loideae, Rosoideae and Dryadoideae, were strongly sup-
ported in the ML tree topology. The genera formerly included
in Rosoideae, Purshia, Cowania, Chamaebatia, Cercocar-
pus and Dryas form a clade, and they are sister to all other
Rosoideae members. This clade corresponds to Dryadoideae,
which was recovered as a sister group to Rosoideae.
Within Rosoideae, six strongly supported (bp = 100,
pp = 1) lineages were resolved corresponding to six tribes:
from the basis Ulmarieae, Colurieae, Rubeae, Agrimonieae,
Roseae and Potentilleae (Fig.1, Online Resources 6–8).
Ulmarieae, Rubeae and Roseae are monogeneric tribes each
with a single genus.
Within Colurieae, Fallugia and Sieversia are basal in the
tribe. The remaining genera are very closely related. Eryth-
rocoma was included in Geum in Flora of North America,
but it is more closely related to Acomastylis. Erythrocoma
nests within Acomastylis, indicating that Erythrocoma
should be merged into Acomastylis rather than to Geum if
Acomastylis was accepted as a distinct genus. A very inclu-
sive Geum is a good choice to include them all even together
There are two strongly supported major subclades within
Agrimonieae. The first subclade comprises Aremonia,
Hagenia, Leucosidea and Spenceria with Spenceria on the
basis. The second subclade comprises nine genera, Acaena,
Bencomia, Cliﬀortia, Marcetella, Margyricarpus, Polylepis,
Sanguisorba, Sarcopoterium and Tetraglochin with Benco-
mia, Marcetella and Sarcopoterium on the basis. Acaena,
Margyricarpus, Polylepis and Tetraglochin are closely
Within Potentilleae, two major subclades were resolved.
The first subclade comprises eight genera: Argentina,
Comarella, Duchesnea, Horkelia, Ivesia, Piletophyllum,
Potentilla and Tylosperma. Very close relationships exist
among Comarella, Horkelia and Ivesia and between Argen-
tina and Piletophyllum. Neither Ivesia nor Potentilla is
monophyletic genus. The second subclade comprises 13
Table 2 Macrofossils of Rosaceae used as calibration points for molecular dating
Anchor fossil Geologic period, Mya Assigned date, Mya Geographic origin References
Rosales crown Cretaceous Fixed (90, 96) Wang etal. (2009)
Cercocarpus myricaefolius Late Eocene 34.07 North America, USA, Colorado MacGinite (1953),
Rosa germerensis Early Eocene 50 North America, USA, Idaho Edelman (1975)
Molecular systematics of Rosoideae (Rosaceae)
Page 5 of 12 9
Fig. 1 Maximum-likelihood phylogram of Rosoideae based on 12 cpDNA regions. Numeric values at nodes indicate maximum parsimony boot-
strap values, maximum likelihood bootstrap values and Bayesian posterior probability support values. *represents 100 or 1.00
X. Chen etal.
9 Page 6 of 12
genera: Alchemilla, Aphanes, Chamaerhodos, Comarum,
Dasiphora, Drymocallis, Fragaria, Lachemilla, Potaninia,
Schistophyllidium, Sibbaldia, Sibbaldianthe and Sibbaldiop-
sis. Very close relationships exist between Dasiphora and
Potaninia, between Schistophyllidium and Sibbaldianthe and
between Sibbaldia and Sibbaldiopsis. Two lineages were
revealed in each of Sibbaldia, Sibbaldiopsis and Dasiphora.
Divergence timesofmajor lineages inRosoideae
The crown Rosaceae diversiﬁed for ca. 89.56 Mya (95%
HPD = 79.58–96.02 Mya) in the Cretaceous (Fig.2). The
crown Rosoideae began to diverge from Dryadoideae ca.
69.77 (95% HPD = 61.28–78.33) Mya in the late Cretaceous.
The divergence times of the tribes of Rosoideae occurred
in the Paleocene and Eocene. Most genera of the subfamily
diversiﬁed in the Miocene.
Reconstruction ofancestral areas
The ancestral area reconstruction within Rosaceae using
S-DIVA suggested at least 398 dispersal, 9 vicariance
and 2 extinction events. Therein, at least 369 dispersal,
8 vicariance and 2 extinction events formed the present
distribution of Rosoideae (Fig.3, Online Resource 9,
Table3). For example, one vicariance event occurred in
the Colurieae crown node, AD → A|D, causing isolated
distribution between North America and Asia. An extinc-
tion event, E → →AD → AD, happened ~ 26 Mya ago in
Alchemilla–Schistophyllidium, explaining their absence in
Europe and thriving in North America and Asia. The sub-
family probably had an origin in North America (37%) and/
or Asia (63%). Five (Ulmarieae, Colurieae, Rubeae, Roseae
and Potentilleae) out of the six tribes most likely evolved in
Asia and North America. Only Agrimonieae appeared to be
from Africa and Asia.
Phylogeny andsubdivision ofRosaceae
The traditional subdivision of Rosaceae into four subfamilies
is not supported by this study, and a modiﬁcation of three
subfamilies, Amygdaloideae, Dryadoideae and Rosoideae, is
supported (Potter etal. 2007). Our results strongly supported
the topology (Amygdaloideae(Dryadoideae,Rosoideae)),
which is congruent with previous studies (Chin etal. 2014;
Zhang etal. 2017), instead of (Rosoideae(Dryadoideae,Sp
iraeoideae = Amygdaloideae)) (Potter etal. 2007) or (Dry
adoideae(Rosoideae,Amygdaloideae) (Xiang etal. 2017).
Our results are consistent with the fact that Dryadoideae
and Rosoideae share some morphological features such as
separation of the hypanthium from the ovary, presence of
stipules and usually achenes as fruits.
Six strongly supported lineages correspond well to six
existing tribes, i.e., Ulmarieae, Colurieae, Rubeae, Agrimo-
nieae, Roseae and Potentilleae from the basal position upon.
Ulmarieae signiﬁcantly diverged from the other tribes, and
its basal systematic position in the subfamily is solid. Dis-
putes may stem from the positions of two monogeneric tribes,
i.e., Rubeae and Roseae. The systematic position of Rubeae
has so far been uncertain. It is resolved in this study to be
between Colurieae and Agrimonieae. This systematic posi-
tion of Rubeae revealed by representative chloroplast genes is
congruent with the phylogeny based on the whole plastomes
(Zhang etal. 2017) but diﬀers from the phylogeny based on
nuclear data, which showed a more basal position than Colu-
rieae (Xiang etal. 2017). Such incongruence between the
topologies based on chloroplast and nuclear genes is prob-
ably due to the nature of sequence data of diﬀerent genomes
and the exact cause is still unknown. The sister relationship
between Roseae and Potentilleae as shown by Potter etal.
(2007) is recovered, but the clade has an extremely short
branch length, suggesting a close relationship among Roseae,
Potentilleae and Agrimonieae as shown by Eriksson etal.
(2003). Support for either topology ((Potentilleae,Roseae)
Sanguisobeae) or ((Potentilleae,Sanguisobeae)Roseae) is not
overwhelming, indicating a close genetic relationship among
them, and factors such as incomplete lineage sorting may
cause phylogenetic uncertainty.
Phylogenetic relationships ofclosely related genera
In the tribes composed of multiple genera, the generic status
of the very closely related genera is questionable. In Colu-
rieae, Acomastylis is paraphyletic because of the existence
of Erythrocoma. Acomastylis 1 forms a clade together with
Erythrocoma, the same relationship shown by Smedmark
and Eriksson (Smedmark and Eriksson 2002). There are
several similar examples in Potentilleae.
Potentilla is a genus of extensive disputes on its circum-
scription. Broadly deﬁned Potentilla includes (1) Dasiphora,
(2) Drymocallis, (3) Comarella–Ivesia–Horkelia–Horke-
liella and (4) Duchesnea. Our results do not support a close
relationship between Alchemilla and Potentilla (Eriksson
etal. 2003). Considering that the members of the broadly
deﬁned Potentilla do not fall into the same clade, such an
all-inclusive genus is not acceptable. Potentilla is better
used in its strict sense, and Dasiphora (including Potaninia)
and Drymocallis are better considered as distinct genera.
Because of the very close relationship among Comarella,
Horkelia, Horkeliella, Ivesia, Duchesnea and Potentilla,
they should be merged into one genus.
Molecular systematics of Rosoideae (Rosaceae)
Page 7 of 12 9
Fig. 2 BEAST chronogram of Rosaceae based on partitioned analysis of 12 cpDNA regions. Plio Pliocene; Quat Quaternary
X. Chen etal.
9 Page 8 of 12
Fig. 3 Ancestral area reconstruction of Rosoideae as inferred by S-DIVA. Numbers at the nodes are Bayesian posterior probabilities obtained
from the BEAST tree. Plio Pliocene; Quat Quaternary. *stands for 100 or 1.00
Molecular systematics of Rosoideae (Rosaceae)
Page 9 of 12 9
Sibbaldia and Sibbaldiopsis are reciprocally non-mono-
phyletic. Although our samples of Sibbaldia only cover par-
tial species in this genus, it is clear that Sibbaldiopsis is very
closely related to Sibbaldia as also shown by Eriksson etal.
(2014). The species in Sibbaldiopsis had been transferred to
Sibbaldia (Paule and Soják 2009).
An obvious diﬀerence of the clade comprising Alche-
milla, Aphanes and Lachemilla from other such clades is
the relatively long branches. The clade is highly supported.
The long branches are probably due to accelerated evolu-
tion of this clade. Although the monophyly of each genus is
retained, a wider circumscription of Alchemilla to include
Aphanes and Lachemilla was suggested (Gehrke etal. 2008).
Considering the long branches leading to each genus, these
genera should be retained.
Chloroplast versusnuclear gene‑based phylogenies
Chloroplast genes have been more extensively used in plant
phylogenetic reconstructions than nuclear genes with the
exception of nuclear ITS. Although the results are generally
congruent between the two molecular data sources, conﬂicts
are often reported. Recently, the phylogeny of Rosaceae
was rebuilt using nuclear genes isolated by sequencing
transcriptomes (Xiang etal. 2017). Although the general
phylogenetic patterns at the subfamilial and tribal levels are
the same, there are a few remarkable diﬀerences in the tree
topologies based on chloroplast genes and nuclear genes. (1)
Systematic position of Dryadoideae. The chloroplast genes
suggest a sister relationship of Dryadoideae to Rosoideae
(Chin etal. 2014; Zhang etal. 2017), but the nuclear genes
indicate a basal position in the family. From the viewpoint
of morphology, the former scenario appears more natural
because the genera in Dryadoideae were the members of
Rosoideae. (2) Systematic position of Rubeae. As discussed
earlier, the chloroplast genes suggest a position between
Colurieae and Agrimonieae, but the nuclear genes indicate
an earlier divergence than Colurieae. (3) Systematic posi-
tion of Roseae. The chloroplast genes indicate a slightly
closer relationship between Roseae and Potentilleae, but the
nuclear genes indicate an earlier divergence of Roseae than
both Agrimonieae and Potentilleae. (4) Systematic position
of some genera. Diﬀerent relationships were revealed by
chloroplast and nuclear genes among Coluria, Taihangia and
Waldsteinia in Colurieae, among Aremonia, Hagenia, Leu-
cosidea, Sanguisorba and Sarcopoterium in Agrimonieae,
and among Alchemilla, Comarus and Fragaria in Potentil-
leae. The problematic lineages or clades often showed short
ancestral branches, but this was not the case in the nuclear
gene-based tree (Xiang etal. 2017, Fig.2). More single
copy nuclear genes should be used to resolve the systematic
uncertainties of the tribes as well as genetic relationships
among genera within tribes.
Origin, expansion anddiversication ofRosoideae
Molecular dating and historical biogeography analyses sug-
gest origins and rapid diversiﬁcation of Rosoideae in Asia
and North America in the Cretaceous. The ages of Rosaceae
and its three subfamilies were estimated to be in the middle
Cretaceous (89.56–82.42 Mya), similar to the earlier estima-
tions (Chin etal. 2014; Zhang etal. 2017). The divergence
time among the six tribes of Rosoideae was inferred to be
between the late Cretaceous to the middle Eocene when the
earth was warmer and moister and tropical forests occurred
worldwide (Zachos etal. 2001). The crown ages of these six
tribes were indicated to be 45.02–10.42 Mya (the middle
Eocene to the late Miocene).
Based on our S-DIVAs, the most recent common ances-
tor (MRCA) for the Rosoideae appeared in Asia and North
America in the late Cretaceous (61.28 to 78.33 Mya), pos-
sibly after the vicariance event. From 55 to 50 Mya, there
was a signiﬁcant global warming episode known as the
Paleocene–Eocene thermal maximum (PETM) and the
early Eocene climatic optima (EECO) events (McInerney
and Wing 2011; Zachos etal. 2001). Climatic change was
likely a driving force for diversiﬁcation in Rosoideae during
the PETM. The uplift of Tibetan Plateau caused by collision
of the Indian plate with the Eurasian plate at ~ 50 Mya cre-
ated considerable new ecological niches which would have
Table 3 Vicariance and
extinction events within
A North America; B South America; C Africa; D Asia; E European
Crown taxon Event Event route Time (Mya)
Leucosidea–Aremonia crown Vicariance CE → C|E 6.72
Margyricarpus–Cliﬀortia crown Vicariance BC → B|C 12.92
Horkelia–Potentilla crown Vicariance AD → A|D 13.57
Alchemilla–Schistophyllidium crown Vicariance/extinction E → →AD → AD 25.86
Leucosidea–Spenceria crown Vicariance CD → C|D 27.11
Alchemilla–Sibbaidia crown Vicariance/extinction AD → A→AE → A|E 28.03
Colurieae crown Vicariance AD → A|D 44.94
Rosoideae–Dryadoideae crown Vicariance AD → A|D 82.42
X. Chen etal.
9 Page 10 of 12
contributed to strong vicariance force in shaping the diver-
siﬁcation process. The fossil record demonstrates a major
diversiﬁcation of Rosaceae during the Eocene (DeVore and
Pigg 2007). A continuous forest belt developed from Asia to
North America across the Beringia known as the boreotropi-
cal ﬂora (Lavin and Luckow 1993; Morley 1999; Tiﬀney
1985a, b; Wolfe 1975; Xiang and Soltis 2001; Zachos etal.
2001), which created habitats for the diversification of
During the icehouse of the Oligocene (34–23 Mya),
the climate became cool and dry, and the warm–temperate
boreotropical forests in the Northern Hemisphere receded.
Species adapted to dry and cold environments thrived in
large parts of North America and Eurasia. The MRCA of
Ulmarieae, Rubeae and Roseae was found within Asia (63%)
and North America (37%) in the early to middle Miocene,
the period when rosaceous lineages diversiﬁed rapidly. Flo-
ristic exchanges between the two continents are thought to
be by the north Atlantic land bridge at an early time and by
the Bering land bridge at a later time (Tiﬀney 1985a, b). The
vicariance could have happened between the populations of
Ulmarieae (or Rubeae or Roseae) in the North America and
Asia due to the break of BLB. The ancestor of these three
tribes were inferred to appear 35–20 Mya in North America
and Asia, when the BLB existed.
Diversiﬁcation of the tribe Potentilleae started at approxi-
mately 45 (38.39–50.12) Mya. The tribe had a widespread
distribution either in Asia (63%) or North America (37%).
The lineage split into two by a dispersal event, and both
had an Asian and North American origin (Dobes and Paule
2010). Similar time (45 Mya) and patterns were postulated
in Colurieae with either Asian and North American (63%)
or North American (37%) origins after a dispersal event.
The MRCA of the tribe Agrimonieae was suggested by
S-DIVA to be of a widespread distribution either between
Africa and Asia (50%) or Africa (50%) approximately 40
(30.42–49.16) Mya after a dispersal event. This was the time
when a global cooling event began at the Eocene–Oligocene
boundary (35 Mya). During this period, the warm-adapted
elements of the boreotropical forests were forced to recede
toward lower latitudes. The climatic change could have
driven ﬂoristic exchanges between Asia and Africa.
Rosaceae is subdivided into three subfamilies, Amygda-
loideae, Dryadoideae and Rosoideae, on the basis of molec-
ular phylogeny. Rosoideae is further subdivided into six
tribes, Ulmarieae, Colurieae, Rubeae, Agrimonieae, Roseae
and Potentilleae. After the recircumscription of some gen-
era, 36 genera are recognized in Rosoideae. It is estimated
that Rosoideae diverged from other subfamilies in the late
Cretaceous some 69.77 (61.28–78.33) Mya, but most genera
diverged in the Miocene. Rosoideae is deduced for having
been thriving in North America and Asia for some 70 Mya.
It reached its current geographical distribution through at
least 369 dispersal, 8 vicariance and 2 extinction events. Of
the 8 vicariance events, three occurred within Potentilleae,
three within Agrimonieae, one in the Colurieae crown node
and one in the Rosoideae crown node. The two extinction
events happened within Potentilleae. These ﬁndings have
signiﬁcantly improved our understanding of the evolution
Acknowledgements We thank Dr. Guojin Zhang and Dr. Bao Nie for
their help in the molecular dating and historical biogeography analy-
ses. This study was partly supported by funds from National Natural
Science Foundation of China (NSFC31872679), Chinese Academy of
Sciences (CAS) Biodiversity Conservation and the Collaborative Inno-
vation Plan, Institute of Forensic Science, Ministry of Public Security,
Compliance with ethical standards
Conflict of interest The authors declare no conﬂict of interests.
Information onElectronic Supplementary Mate-
Online Resource 1. Material list and newly produced sequences in this
Online Resource 2. Primers and procedures used to amplify DNA frag-
Online Resource 3. Taxa and GenBank accessions used in this study.
Online Resource 4. Detailed Fig.2 of chronogram of Rosaceae based
on partition of12 cpDNA regions derived from BEAST. Numbers at
nodes are divergencetime of clades.
Online Resource 5. Geographical distributions of extant species in
Rosoideae used inthis study.
Online Resource 6. Detailed Fig.1 of maximum likelihood phylogram
of Rosoideaebased on 12 cpDNA regions of all taxa. Numeric values
at nodes indicate23maximum likelihood bootstrap values.
Online Resource 7. Detailed Fig.1 of BI tree of Rosoideae based on 12
cpDNAregions of all taxa. Numbers at the nodes are Bayesian poste-
rior probabilitiesand PP values ≥0.50 are shown
Online Resource 8. Detailed Fig. 1 of maximum parsimony tree of
Rosoideae basedon 12 cpDNA regions of all taxa. Numbers at the
nodes are bootstrap values,and BP values ≥50 are shown.
Online Resource 9. Detailed Fig.3 of ancestral area reconstruction of
Rosoideaeinferred by S-DIVA. Numbers at the nodes are Bayesian
posterior probabilitiesobtained from the BEAST tree.
Online Resource 10. Sequence matrix A containing 12 chloroplast re-
gions of 169taxa for phylogenetic analyses.
Online Resource 11. Sequence matrix B containing 12 chloroplast re-
gions of 130taxa for molecular dating and biogeographical analyses.
Chin SW, Shaw J, Haberle R, Wen J, Potter D (2014) Diversiﬁcation
of almonds, peaches, plums and cherries—molecular systematics
Molecular systematics of Rosoideae (Rosaceae)
Page 11 of 12 9
and biogeographic history of Prunus (Rosaceae). Molec Phylogen
Evol 76:34–48. https ://doi.org/10.1016/j.ympev .2014.02.024
DeVore ML, Pigg KB (2007) A brief review of the fossil history of
the family Rosaceae with a focus on the Eocene Okanogan High-
lands of eastern Washington State, USA, and British Columbia,
Canada. Pl Syst Evol 266:45–57. https ://doi.org/10.1007/s0060
Dobes C, Paule J (2010) A comprehensive chloroplast DNA-based
phylogeny of the genus Potentilla (Rosaceae): Implications for
its geographic origin, phylogeography and generic circumscrip-
tion. Molec Phylogen Evol 56:156–175. https ://doi.or g/10.1016/j.
Dong W, Liu J, Yu J, Wang L, Zhou S (2012) Highly variable chlo-
roplast markers for evaluating plant phylogeny at low taxonomic
levels and for DNA barcoding. PLoS ONE 7:e35071. https ://doi.
org/10.1371/journ al.pone.00350 71
Dong W, Cheng T, Li C, Xu C, Long P, Chen C, Zhou S (2014) Dis-
criminating plants using the DNA barcode rbcLb: an appraisal
based on a large data set. Molec Ecol Resources 14:336–343. https
Dong W, Xu C, Li C, Sun J, Zuo Y, Shi S, Cheng T, Guo J, Zhou S
(2015) ycf1, the most promising plastid DNA barcode of land
plants. Sci Rep 5:8348. https ://doi.org/10.1038/srep0 8348
Drummond AJ, Suchard MA, Xie D, Rambaut A (2012) Bayesian phy-
logenetics with BEAUti and the BEAST 1.7. Molec Biol Evol
29:1969–1973. https ://doi.org/10.1093/molbe v/mss07 5
Edelman DW (1975) The Eocene Germer Basin ﬂora of south-central
Idaho. MSc Thesis, University of Idaho, Moscow
Eriksson T, Donoghue MJ, Hibbs MS (1998) Phylogenetic analysis
of Potentilla using DNA sequences of nuclear ribosomal internal
transcribed spacers (ITS), and implications for the classiﬁcation
of Rosoideae (Rosaceae). Pl Syst Evol 211:155–179. https ://doi.
Eriksson T, Hibbs MS, Yoder AD, Delwiche CF, Donoghue MJ (2003)
The phylogeny of Rosoideae (Rosaceae) based on sequences of
the internal transcribed spacers (ITS) of nuclear ribosomal DNA
and the trnL/F region of chloroplast DNA. Int J Pl Sci 164:197–
211. https ://doi.org/10.1086/34616 3
Eriksson T, Lundberg M, Töpel M, Östensson P, Smedmark JEE (2014)
Sibbaldia: a molecular phylogenetic study of a remarkably poly-
phyletic genus in Rosaceae. Pl Syst Evol 301:171–184. https ://doi.
Evanhoﬀ E, Gregory-Wodzicki KM, Johnson KR (2001) Fossil ﬂora
and stratigraphy of the Florissant Formation, Colorado. Denver
Museum of Nature & Science, Denver
Gardens RB (2007) DNA barcoding. Available at:http://www.kew.org/
barco ding/proto cols.html. Accessed 21 Apr 2014
Gehrke B, Brauchler C, Romoleroux K, Lundberg M, Heubl G, Eriks-
son T (2008) Molecular phylogenetics of Alchemilla, Aphanes
and Lachemilla (Rosaceae) inferred from plastid and nuclear
intron and spacer DNA sequences, with comments on generic
classiﬁcation. Molec Phylogen Evol 47:1030–1044. https ://doi.
Hillis DM (1998) Taxonomic sampling, phylogenetic accuracy, and
investigator bias. Syst Biol 47:3–8. https ://doi.org/10.1080/10635
Hutchinson J (1964) The genera of ﬂowering plants. Clarendon Press,
Kalkman C (2004) Rosaceae. In: Kubitzki K (ed) The families and
genera of vascular plants. Springer, Berlin, pp 343–386. https ://
Katoh K, Standley DM (2013) MAFFT multiple sequence alignment
software version 7: improvements in performance and usability.
Molec Biol Evol 30:772–780. https ://doi.org/10.1093/molbe v/
Lavin M, Luckow M (1993) Origins and relationships of tropical North
America in the context of the boreotropics hypothesis. Amer J
Bot 80:1–14. https ://doi.org/10.1002/j.1537-2197.1993.tb137 61.x
Li H-L (1952) Floristic relationships between eastern Asia and east-
ern North. Trans Amer Philos Soc 42:371–429. https ://doi.
Li J, Wang S, Yu J, Wang L, Zhou S (2013) A modiﬁed CTAB protocol
for Plant DNA extraction. Chin Bull Bot 48:72–78. https ://doi.
Librado P, Rozas J (2009) DnaSP v5: a software for comprehensive
analysis of DNA polymorphism data. Bioinformatics 25:1451–
1452. https ://doi.org/10.1093/bioin forma tics/btp18 7
Lu L, Gu C, Li C, Alexander C, Bartholomew B, Brach AR, Bouf-
ford DE, Ikeda H, Ohba H, Robertson KR, Spongberg SA (2003)
Rosaceae. In: Wu Z, Raven PH, Hong D (eds) Flora of China.
Science Press/Missouri Botanical Garden Press, Beijing/St.Louis,
MacGinite HD (1953) Fossil plants of the Florissant beds of Colorado.
Carnegie Institute of Washington, Washington, DC
McInerney FA, Wing SL (2011) The Paleocene-Eocene Thermal Maxi-
mum: a perturbation of carbon cycle, climate, and biosphere with
implications for the future. Annual Rev Earth Planet Sci 39:489–
516. https ://doi.org/10.1146/annur ev-earth -04061 0-13343 1
Miller MA, Pfeiﬀer W, Schwartz T (2010) Creating the CIPRES Sci-
ence Gateway for inference of large phylogenetic trees. In: 2010
Gateway Computing Environments Workshop (GCE). IEEE, New
Orleans, pp 1–8
Morgan DR, Soltis DE, Robertson KR (1994) Systematic and evolution-
ary implications of rbcL sequence variation in Rosaceae. Amer J
Bot 81:890–903. https ://doi.org/10.1002/j.1537-2197.1994.tb155
Morley RJ (1999) Origin and evolution of tropical rain forests. Wiley,
Paule J, Soják J (2009) Taxonomic comments on the genus Sibbaldiop-
sis Rydb. (Rosaceae). J Natl Mus (Pragus) Nat Hist Ser 178:15–16
Posada D, Buckley TR (2004) Model selection and model averaging
in phylogenetics: advantages of Akaike information criterion and
Bayesian approaches over likelihood ratio tests. Syst Biol 53:793–
808. https ://doi.org/10.1080/10635 15049 05223 04
Posada D, Crandall KA (1998) MODELTEST: testing the model
of DNA substitution. Bioinformatics 14:817–818. https ://doi.
org/10.1093/bioin forma tics/14.9.817
Potter D, Gao F, Bortiri PE, Oh SH, Baggett S (2002) Phylogenetic
relationships in Rosaceae inferred from chloroplast matK and
trnL-trnF nucleotide sequence data. Pl Syst Evol 231:77–89. https
://doi.org/10.1007/s0060 60200 012
Potter D, Eriksson T, Evans RC, Oh S, Smedmark JEE, Morgan DR,
Kerr M, Robertson KR, Arsenault M, Dickinson TA, Campbell
CS (2007) Phylogeny and classiﬁcation of Rosaceae. Pl Syst Evol
266:5–43. https ://doi.org/10.1007/s0060 6-007-0539-9
Rambaut A (2002) Se-Al, sequence alignment editor v2.0a ll [online].
Available at:http://tree.bio.ed.ac.uk/softw are/seal/
Rambaut A, Drummond AJ (2009) Tracer: MCMC trace analysis tool,
version 1.5.0. Available at:http://tree.bio.ed.ac.uk/softw are/trace
r/. Accessed 19 May 2014
Rambaut A, Drummond AJ (2010a) LogCombiner: MCMC output
combiner, version 1.5.4. University of Edinburgh, Institute of
Rambaut A, Drummond AJ (2010b) TreeAnnotator: MCMC output
analysis, version 1.5.4.University of Edinburgh, Institute of Evo-
Raven PH (1972) Plant species disjunctions: a summary. Ann Missouri
Bot Gard 59:234–246. https ://doi.org/10.2307/23947 56
Ronquist F, Huelsenbeck JP (2003) MrBayes 3: Bayesian phylogenetic
inference under mixed models. Bioinformatics 19:1572–1574.
https ://doi.org/10.1093/bioin forma tics/btg18 0
X. Chen etal.
9 Page 12 of 12
Schulze-Menz GK (1964) Rosaceae. In: Melchior H (ed) In Engler’s
Syllabus der Pﬂanzenfamilien Gebruder. Borntraeger, Berlin, pp
Smedmark JEE, Eriksson T (2002) Phylogenetic relationships
of Geum (Rosaceae) and relatives inferred from the nrITS
and trnL-trnF regions. Syst Bot 27:303–317. https ://doi.
Stamatakis A (2014) RAxML version 8: a tool for phylogenetic
analysis and post-analysis of large phylogenies. Bioinformatics
30:1312–1313. https ://doi.org/10.1093/bioin forma tics/btu03 3
Swoﬀord DL (2002) PAUP*: phylogenetic analysis using parsimony (*
and other methods). Version 4. Sinauer Associates, Sunderland.
https ://doi.org/10.1002/04716 50129 .dob05 22
Taberlet P, Gielly L, Pautou G, Bouvet J (1991) Universal primers for
ampliﬁcation of three non-coding regions of chloroplast DNA. Pl
Molec Biol 17:1105–1109. https ://doi.org/10.1007/BF000 37152
Takhtajan A (1997) Diversity and classiﬁcation of ﬂowering plants.
Columbia University Press, New York
Thorne RF (1972) Major disjunctions in the geographic ranges
of seed plants. Quatern Rev Biol 47:365–411. https ://doi.
Tiﬀney BH (1985a) The Eocene North Atlantic land bridge: its impor-
tance in Tertiary and modern phytogeography of the Northern
Hemisphere. J Arnold Arbor 66:243–273. https ://doi.org/10.5962/
Tiﬀney BH (1985b) Perspectives on the origin of the ﬂoristic similarity
between eastern Asia and eastern North America. J Arnold Arbor
66:73–94. https ://doi.org/10.5962/bhl.part.13179
Tiﬀney BH, Manchester SR (2001) The use of geological and paleon-
tological evidence in evaluating plant phylogeographic hypotheses
in the Northern Hemisphere Tertiary. Int J Pl Sci 162:S3–S17.
https ://doi.org/10.1086/32388 0
Wang H, Moore MJ, Soltis PS, Bell CD, Brockington SF, Alexandre
R, Davis CC, Latvis M, Manchester SR, Soltis DE (2009) Rosid
radiation and the rapid rise of angiosperm-dominated forests.
Proc Natl Acad Sci USA 106:3853–3858. https ://doi.org/10.1073/
Wen J (1998) Evolution of the eastern Asian and eastern North Ameri-
can disjunct pattern: insights from phylogenetic studies. Korean
J Pl Tax 28:63–81. https ://doi.org/10.11110 /kjpt.1998.28.1.063
Wen J (1999) Evolution of eastern Asian and eastern North Amer-
ica disjunct pattern in ﬂowering plants. Annual Rev Ecol Syst
30:421–455. https ://doi.org/10.2307/22169 1
Wen J, Ickert-Bond S, Nie ZL, Li R (2010) Timing and modes of evolu-
tion of eastern Asian—North America biogeographic disjunctions
in seed plants. In: Long M, Gu H, Zhou Z (eds) Darwin’s heritage
today—proceedings of the Darwin 200 Beijing international con-
ference. Higher Education Press, Beijing, pp 252–269
Wolfe JA (1975) Some aspects of plant geography of the Northern
hemisphere during the late cretaceous and tertiary. Ann Missouri
Bot Gard 62:264–279. https ://doi.org/10.2307/23951 98
Xiang Q-Y, Soltis DE (2001) Dispersal-vicariance analyses of inter-
continental disjuncts: historical biogeographical implications for
angiosperms in the Northern hemisphere. Int J Pl Sci 162:S29–
S39. https ://doi.org/10.1086/32333 2
Xiang QY, Soltis DE, Soltis PS, Manchester SR, Crawford DJ (2000)
Timing the eastern Asian–eastern North American floristic
disjunction: molecular clock corroborates paleontological esti-
mates. Molec Phylogen Evol 15:462–472. https ://doi.or g/10.1006/
Xiang Y, Huang CH, Hu Y, Wen J, Li S, Yi T, Chen H, Xiang J,
Ma H (2017) Evolution of Rosaceae fruit types based on nuclear
phylogeny in the context of geological times and genome duplica-
tion. Molec Biol Evol 34:262–281. https ://doi.org/10.1093/molbe
Yu Y, Harris AJ, He X (2010) S-DIVA (statistical dispersal-vicari-
ance analysis): a tool for inferring biogeographic histories. Molec
Phylogen Evol 56:848–850. https ://doi.org/10.1016/j.ympev
Yu J, Xue J-H, Zhou S-L (2011) New universal matK primers for DNA
barcoding angiosperms. J Syst Evol 49:176–181. https ://doi.org/1
Yu Y, Harris AJ, Blair C, He X (2015) RASP (Reconstruct Ances-
tral State in Phylogenies): a tool for historical biogeography.
Molec Phylogen Evol 87:46–49. https ://doi.org/10.1016/j.ympev
Zachos J, Pagani M, Sloan L, Thomas E, Billups K (2001) Trends,
rhythms, and aberrations in global climate 65Ma to present. Sci-
ence 292:686–693. https ://doi.org/10.1126/scien ce.10594 12
Zhang SD, Jin JJ, Chen SY, Chase MW, Soltis DE, Li HT, Yang JB, Li
DZ, Yi TS (2017) Diversiﬁcation of Rosaceae since the Late Cre-
taceous based on plastid phylogenomics. New Phytol 214:1355–
1367. https ://doi.org/10.1111/nph.14461
Publisher’s Note Springer Nature remains neutral with regard to
jurisdictional claims in published maps and institutional aﬃliations.