ArticlePDF Available

Phylogenomic analyses reveal extensive gene flow within the magic flowers ( Achimenes )


Abstract and Figures

Premise of the Study The Neotropical Gesneriaceae is a lineage known for its colorful and diverse flowers, as well as an extensive history of intra‐ and intergeneric hybridization, particularly among Achimenes (the magic flowers) and other members of subtribe Gloxiniinae. Despite numerous studies seeking to elucidate the evolutionary relationships of these lineages, relatively few have sought to infer specific patterns of gene flow despite evidence of widespread hybridization. Methods To explore the utility of phylogenomic data for reassessing phylogenetic relationships and inferring patterns of gene flow among species of Achimenes, we sequenced 12 transcriptomes. We used a variety of methods to infer the species tree, examine gene tree discordance, and infer patterns of gene flow. Key Results Phylogenomic analyses resolve clade relationships at the crown of the lineage with strong support. In contrast to previous analyses, we recovered strong support for several new relationships despite a significant amount of gene tree discordance. We present evidence for at least two introgression events between two species pairs that share pollinators, and suggest that the species status of Achimenes admirabilis be reexamined. Conclusions Our study demonstrates the utility of transcriptome data for phylogenomic analyses, and inferring patterns of gene flow despite gene tree discordance. Moreover, these data provide another example of prevalent interspecific gene flow among Neotropical plants that share pollinators.
Content may be subject to copyright.
American Journal of Botany 105(4): 1–15, 2018; © 2018 Botanical Society of America 1
Botanists have long suspected a signicant role for hybridization
in diversication processes (Anderson, 1949; Stebbins, 1950; Grant,
1971), where it serves as both a direct source of novel lineages
through hybrid speciation and a potential source of new variation
through introgression. Molecular studies from across the tree of life
have both conrmed the hybrid origin of many organisms (such as
wheat and corn) and have revealed a more extensive history of hy-
bridization than previously documented (Mallet etal., 2016; Roda
etal., 2017; Vargas etal., 2017). Interspecic gene ow may con-
tribute to lineage convergence because of introgression, occurring
more rapidly among closely related species and resulting in an in-
crease in reproductive isolation with genetic divergence (Coyne and
Orr, 1997; Moyle etal., 2004; Scopece et al., 2007). Alternatively,
hybridization may promote rapid reproductive isolation of sym-
patric species due to selection against the formation of unt hy-
brid ospring (‘reinforcement’), thereby accelerating the process
of speciation (Servedio and Noor, 2003). Furthermore, favorable
new gene combinations may give rise to new species, which may
become instantly reproductively isolated from both parental species
(Vereecken etal., 2010). ere is a general lack of knowledge about
the commonality of dierent pathways to interspecic gene ow,
and it is still uncertain whether hybrids contain approximately equal
contributions from each parent or whether hybridization events in-
volve asymmetrical contributions. Newly developed approaches to
obtain genome- wide estimates of phylogenetic history will consid-
erably improve our understanding of gene ow in plants. We there-
fore implemented a transcriptome- sequencing based approach to
infer phylogenomic patterns of gene ow within Achimenes—one
of the most colorful and diverse genera of Central American plants.
Achimenes, commonly known as the “magic owers,” is a genus
in the large African violet family (Gesneriaceae) and is well known
for its oral diversity among closely related species (Fig.1; Ramírez
Roa, 1987; Roalson etal., 2003; Roberts and Roalson, 2017). Recent
divergence time estimates and biogeographic reconstructions have
Phylogenomic analyses reveal extensive gene ow within
the magic owers (Achimenes)
Wade R. Roberts1,2,3 and Eric H. Roalson1,2
Manuscript received 20 September 2017; revision accepted
2February 2018.
1 Molecular Plant Sciences Graduate Program,Washington State
University, Pullman, Washington 99164-1030, USA
2 School of Biological Sciences,Washington State University,
Pullman, Washington 99164-4236, USA
3 Author for correspondence (e-mail:
Citation: Roberts, W. R. and E. H. Roalson. 2018. Phylogenomic
analyses reveal extensive gene ow within the magic owers
(Achimenes). American Journal of Botany 105(4): 1–15.
PREMISE OF THE STUDY: The Neotropical Gesneriaceae is a lineage known for its colorful
and diverse owers, as well as an extensive history of intra- and intergeneric hybridization,
particularly among Achimenes (the magic owers) and other members of subtribe
Gloxiniinae. Despite numerous studies seeking to elucidate the evolutionary relationships
of these lineages, relatively few have sought to infer specic patterns of gene ow despite
evidence of widespread hybridization.
METHODS: To explore the utility of phylogenomic data for reassessing phylogenetic
relationships and inferring patterns of gene ow among species of Achimenes, we sequenced
12 transcriptomes. We used a variety of methods to infer the species tree, examine gene tree
discordance, and infer patterns of gene ow.
KEY RESULTS: Phylogenomic analyses resolve clade relationships at the crown of the
lineage with strong support. In contrast to previous analyses, we recovered strong support
for several new relationships despite a signicant amount of gene tree discordance. We
present evidence for at least two introgression events between two species pairs that share
pollinators, and suggest that the species status of Achimenes admirabilis be reexamined.
CONCLUSIONS: Our study demonstrates the utility of transcriptome data for phylogenomic
analyses, and inferring patterns of gene ow despite gene tree discordance. Moreover, these
data provide another example of prevalent interspecic gene ow among Neotropical plants
that share pollinators.
KEY WORDS Achimenes; gene ow; Gesneriaceae; hybridization; introgression; phylogenetic
networks; phylogenomics; transcriptomics.
2 American Journal of Botany
indicated a Central American origin for Achimenes
between 7.7 and 14.2 Mya (Roalson and Roberts,
2016). e genus comprises 26 species found
throughout Mexico and Central America, with the
center of diversity in central and southern Mexico
(Fig.2A, B; Ramírez Roa, 1987). Achimenes contains
a mixture of widely distributed species, such as A.
grandiflora (Schltdl.) DC. and A. longiflora DC.,
and many narrow endemics, such as A. admirabilis
Wiehler and A. cettoana H.E. Moore. Many species
overlap in distribution and are oen found growing
in sympatry. Most species of the genus principally
inhabit oak and pine forests, and generally are found
in seasonally dry areas where the production of scaly
rhizomes and propagules allows for seasonal dor-
mancy (Ramírez Roa, 1987). ese scaly rhizomes
characterize this genus and others in the subtribe
Gloxiniinae and oer an important feature useful
for the production and cultivation of these plants.
Enormous diversity in oral form among the
many closely related Achimenes species represents
a feature thought to be associated with speciation
(Fig. 1; Ramírez Roa, 1987; Roalson et al., 2003).
Evidence from phylogenetic, morphological, and
ecological studies oers extensive support for polli-
nation syndromes in Neotropical gesneriads (Perret
etal., 2007; Martén- Rodríguez etal., 2009; Martén-
Rodríguez etal., 2010; Roalson and Roberts, 2016).
In Achimenes, species are divided into those with
hummingbird- , buttery- , bee- , and female eugloss-
ine bee- pollination (Ramírez Roa, 1987; Roalson
etal., 2003). e owers within each of these groups
share similar characters. For example, those with
hummingbird- pollination tend to have red, tubu-
lar owers with ample nectar provided as a reward,
while those with bee- pollination are small, white,
funnelform owers. e genetic basis for transitions
in oral form, including traits such as ower color,
thought to be closely associated with pollinators,
has been studied extensively in many plant systems
(Hoballah etal., 2007; Des Marais and Rausher, 2010;
Wessinger and Rausher, 2014; Roberts and Roalson,
2017). Along with shared distribution, habitat pref-
erences, and elevational ranges among Achimenes,
the high level of oral divergence among closely re-
lated species in sympatry might argue for speciation
being driven by mechanisms such as pollination,
hybridization, or time of owering (Wiehler, 1983).
Achimenes has a rich history of horticultural inter-
est because of the diversity of owers found among its
members. Interest in these plants peaked during the
Victorian Era with many new species being identied
in Mexico and brought into cultivation, where many
FIGURE 1. Flowers of the Achimenes sampled in the
current study. (A) A. admirabilis, (B) A. antirrhina, (C) A.
candida, (D) A. cettoana, (E) A. erecta, (F) A. grandiora,
(G) A. longiora, (H) A. misera, (I) A. patens ‘Major’ and (J)
A. pedunculata. All photos by Wade R. Roberts.
2018, Volume 105 Roberts and Roalson—Phylogenomics in Achimenes3
of the earliest hybrid varieties were exhibited at horticultural shows
(Gordon, 1846; Moore, 1859). e extraordinary range of desirable
colors and shapes in Achimenes has provided many growers material
to produce well over 200 dierent hybrids and varieties (Becker, 2008).
In addition to the abundance of horticultural hybrids, numerous nat-
ural interspecic hybrids among populations of Achimenes in Mexico
are also known (Wiehler, 1983). Specimens can oen display mixed
characteristics with other species found in sympatry, particularly with
species that share similar pollinators. Given the ease with which hy-
brids can be produced among Achimenes, it is likely that gene ow has
occurred more than once during the history of this group.
Only a few attempts have been made to reconstruct a molecu-
lar phylogeny of Achimenes (Roalson etal., 2003) and no studies
have tested specic hypotheses about the role of hybridization in
FIGURE2. Distribution and phylogenetic relationships of Achimenes. Range distributions of the species sampled in (A) Clade 1 and (B) Clade 2 of
Achimenes throughout Mexico and Central America. Location data was downloaded from the Global Biodiversity Information Facility (
(C) Phylogram summarizing results from phylogenetic analyses of Achimenes transcriptomes. Gesneria cuneifolia and Eucodonia verticillata were used
as outgroups. Analyses of concatenated sequences and gene trees converged on the same topology. Branch length units are substitutions per site and
are derived from maximum likelihood analysis in RAxML. Numbers at nodes: bootstrap support from RAxML, ASTRAL, and ASTRID. Bootstrap support
of 100 is labeled as ‘+’, and support otherwise is listed numerically.
4 American Journal of Botany
aecting the evolutionary history of the lineage. Gene ow between
closely related, or more distantly related, species oen results in a
strong conict between gene trees with the traditional hierarchical
(bifurcating) representation of species and is best represented by a
reticulation network (McBreen and Lockhart, 2006; Huson etal.,
2011; Solís- Lemus and Ané, 2016). Reconstruction of phylogenetic
relationships using genome- wide data can be problematic because
hybridization is a major cause of topological incongruence between
gene trees (McBreen and Lockhart, 2006; Folk etal., 2017). Studying
these incongruences oers an opportunity to detect hybrid specia-
tion. However, confounding population genetic processes such as
lineage sorting might mislead inference of the real contribution of
hybridization to the observed patterns of gene tree incongruence
(Linder and Rieseberg, 2004; Kubatko, 2009; Goulet etal., 2017).
is is especially common among closely related species where lin-
eage sorting is not complete, leading to nonmonophyletic species
assemblages (Schmidt- Lebuhn etal., 2012).
Sampling exemplars of 10 species, including several of the most
broadly distributed species and several narrow endemics, our initial
goal was to reevaluate the phylogenetic hypothesis of Roalson etal.
(2003) using thousands of loci derived from transcriptome sequenc-
ing. Recent advancements in methods used to detect gene ow using
genomic data now oer new opportunities to understand the process
of speciation (Joly etal., 2009; Martin etal., 2015; Rosenzweig etal.,
2016). Transcriptome sequencing provides excellent resources for
evolutionary studies, but has so far been little used for studies of gene
ow and speciation (Roda etal., 2017). Here, we employ transcrip-
tome sequencing from across 10 species of Achimenes in three com-
plementary strategies to uncover phylogenetic patterns and detect
the presence and direction of gene ow. We hypothesized that some
level of gene ow occurred within this group in the past, particularly
between A. grandiflora and A. patens Benth., given their close mor-
phological similarities and sympatric range. First, we used gene trees
to estimate phylogenetic networks, which allow reticulation events
such as hybridization (Solís- Lemus and Ané, 2016). Second, we in-
vestigated patterns of divergence at multiple informative sites using
the D- statistic (also known as the ‘ABBA- BABA’ test [Green etal.,
2010]). Finally, we investigated variation in gene ow by compar-
ing the relative divergence of genes with discordant gene trees across
multiple samples. Our results suggest instances of gene ow and po-
tential introgression among several species of Achimenes, while also
calling into question the species status of A. admirabilis.
Taxon sampling and tissue collection
We sampled 10 of the 26 currently recognized species of Achimenes,
including the entirety of Clade 1 and a subset of Clade 2 (sensu
Roalson etal., 2003; Fig.1; Appendix1). We sampled 5 of 13 spe-
cies in Clade 2, which included the entire Clade 2a of Roalson etal.
(2003) as well as A. patens. ose species in Clade 2 not included
in the current study are ones that are highly endemic in southern
Mexico and not in cultivation. Each of these clades contain some
of the most widespread species (e.g., A. longiflora) and some nar-
row endemic species (e.g., A. admirabilis; Figs.2A, B). ese 10
species were chosen based on their hypothesized close relation-
ships and their diversity in oral form and pollination syndrome.
Additionally, we used the variety A. patens ‘Major’ as our exemplar
of A. patens. On the basis of previous molecular work with Sanger-
sequenced loci (Roalson etal., 2008; Roalson and Roberts, 2016),
two species were chosen as outgroups to represent related lineages
in the Gesnerieae: Eucodonia verticillata (M. Martens & Galeotti)
Wiehler and Gesneria cuneifolia (DC.) Fritsch (Appendix1). Plants
for all sampled species were grown in standard greenhouse con-
ditions, under 16 h days, 24–27°C, and 80–85% humidity. Flower
buds from two developmental stages were sampled: an immature
bud stage (Bud) and an intermediate stage (D) pre- anthesis. ree
biological replicates (accessions) each were sampled for both Bud
and D stages in all species, contributing to a total of six samples
for each species. is sampling scheme will allow additional studies
to examine gene expression during development as it relates to the
evolution of oral form and pollination syndrome. Tissue was col-
lected from plants and immediately ash frozen in liquid nitrogen.
Total RNA was extracted from each frozen sample using an RNEasy
Plant Kit (Qiagen, Valencia, California, USA) according to the man-
ufacturer’s directions. e quality and quantity of the RNA samples
were assessed with 1.0% agarose gels and the Fragment Analyzer
(Advanced Analytical Technologies, Ankeny, Iowa, USA) at the
Washington State University Biotechnology Core Lab (Pullman,
Washington, USA). e RNA samples having 28S/18S rRNA ra-
tios approximately 2:1 and RNA Quality Numbers (RQN) ≥8 were
found to be high quality and were used for library preparation.
Library preparation and sequencing
Ribosomal depleted RNA samples were prepared using the RiboMinus
Plant Kit (ermo Fisher Scientic, Waltham, Massachusetts, USA),
using ≤10 μg total RNA as input, followed by an ethanol precipita-
tion to concentrate the RNA and ensure recovery of smaller (<200
nt, nucleotides) RNA. e ethanol precipitation was performed by
adding the following to the ribosomal depleted RNA: 1 μl glycogen
(20 μg/μl; ermo Fisher Scientic, Waltham, Massachusetts, USA),
1/10 sample volume of 3M sodium acetate, and 2.5× sample volume
of 100% ethanol. Samples were then incubated at –80°C for 1 hour,
followed by centrifuging for 15 min at ≥12,000 × g, and washing
twice with 70% ethanol and centrifuging for 5 min at ≥12,000 × g.
e supernatant was removed aer each centrifugation step. e re-
sulting ribosomal depleted RNA was eluted into 30 μl nuclease- free
water and was quantied using a Qubit 2.0 Fluorometer (RNA HS
assays; ermo Fisher Scientic, Waltham, Massachusetts, USA).
We prepared stranded RNA libraries for the 64 samples using
the NEBNext Ultra Directional RNA Library Kit (New England
Biolabs, Ipswich, Massachusetts, USA). For each sample, 10 ng of
eluted RNA was fragmented to 400 nt and primed using 1 μl of ran-
dom hexamers and 4 μl of NEBNext First Strand Synthesis Reaction
Buer (New England Biolabs, Ipswich, Massachusetts, USA), incu-
bating at 94° C for 15 min. en, we performed rst- strand cDNA
synthesis by combining the fragmented RNA with 0.5 μl Murine
RNase Inhibitor (New England Biolabs, Ipswich, Massachusetts,
USA), 1 μl of ProtoScript II Reverse Transcriptase (New England
Biolabs, Ipswich, Massachusetts, USA), and 5 μl of Actinomycin D
(0.1 μg/μl; Sigma- Aldrich, St. Louis, Missouri, USA). e solutions
were incubated for 10 min at 25°C, 15 min at 42°C, and 15 min at
70°C. For second- strand synthesis, we added 8 μl of NEBNext Second
Strand Synthesis Reaction Buer (New England Biolabs, Ipswich,
Massachusetts, USA) and 4 μl of NEBNext Second Strand Synthesis
Enzyme Mix (New England Biolabs, Ipswich, Massachusetts, USA).
e reaction was incubated at 16°C for 1 hour. Aer second- strand
2018, Volume 105 Roberts and Roalson—Phylogenomics in Achimenes5
synthesis, the reaction was cleaned twice up using 1.8X Agencourt
AMPure XP beads (Beckman Coulter, Indianapolis, Indiana, USA)
and eluted into 55.5 μl of nuclease- free H2O. e double- stranded
cDNA was input for end- repair, dA- tailing, and adapter ligation
with Illumina TruSeq barcoded adapters. e ligation reaction
was puried using 1.0X AMPure XP beads and eluted into 17 μl of
nuclease- free H2O. For the library amplication reaction, we ran
the initial denaturation at 98°C for 30 s, followed by 12 cycles of
denaturation at 98° C for 10 s, annealing and extension at 65°C for
75 s, and nal extension at 65°C for 5 min.
Prepared libraries were checked for quality and quantity us-
ing three methods: the Qubit (dsDNA HS assays; ermo Fisher
Scientic Waltham, Massachusetts, USA), the Agilent BioAnalyzer
2100 (High Sensitivity DNA Kit; Agilent Technologies, Santa
Clara, California, USA), and qPCR (KAPA Library Quantication
Kit; Kapa Biosystems, Wilmington, Massachusetts, USA) per-
formed on a Bio- Rad CFX96 Touch Real- Time PCR machine
(Bio- Rad Laboratories, Hercules, California, USA). ese assays
measured library concentration (ng/μl) and library molarity (nM/l;
BioAnalyzer and qPCR only) to create library pools. Libraries were
pooled based on nanomolar concentrations and puried once using
0.6X AMPure XP beads. e library pool was sequenced on one
lane of an Illumina HiSeq2500 at the Washington State University
Genomics Core Lab (Spokane, Washington, USA) for paired- end
101 bp reads. All sequence data were deposited in the Sequence
Read Archive on GenBank (BioProject: PRJNA401042).
Transcriptome assembly and homology inference
e per base quality of the raw reads, adapter content, and per base
sequence content was rst assessed for each sample using the tools
implemented in FastQC (Andrews, 2010) to determine if any samples
returned poor quality reads. Trimmomatic v0.36 (Bolger etal. 2014)
was then used to lter low- quality paired- end sequence reads, per-
forming the following steps in this order: removing Illumina adapters
(TruSeq3- PE- 2.fa le provided by Trimmomatic), removing the rst
13 bases of each read, removing leading and trailing low quality or N
bases (below quality 3), scanning each read with a 4- base wide slid-
ing window and cutting when average quality per base drops below
15, and removing reads that are less than 50 bases long aer these
steps (ILLUMINACLIP:TruSeq3- PE2.fa:2:30:10 HEADCROP:13
e rst 13 bases at the 5- end from each read were trimmed based
on the per base sequence content data from FastQC. e use of ran-
dom hexamers during Illumina library preparation can create biases
in nucleotide composition, aecting the uniformity of the read loca-
tion along expressed transcripts (Hanson etal., 2010). Filtered reads
generated in the current study were combined with the sequence data
generated from Roberts and Roalson (2017). Reads were de novo as-
sembled into contigs using Trinity v2.3.0 (Grabherr etal., 2011; Haas
etal., 2013). We generated reference transcriptomes for each species
by using all samples derived from an individual species. For assem-
bly, we used default settings in Trinity and specied reverse- forward
read alignment (–SS_lib_type RF). ese assemblies were not ltered
using expression levels (FPKM or TPM) because our downstream
analyses are not considering gene expression and we did not wish to
discard any biologically relevant transcripts from our data. We pre-
dicted open reading frames (ORF) and identied coding sequences
(CDS) and peptide sequences for each transcriptome assembly with
TransDecoder v3.0.1 (Haas etal., 2013), using a BLASTp search against
the SwissProt database ( to increase the sensitivity
of functional coding region identication. While TransDecoder may
include 5- and 3- untranslated regions (UTRs) in its output, we used
both coding sequence (CDS) and peptide sequences for all further
analyses. Transcriptome sequence redundancy was additionally re-
duced with CDHIT- EST v4.6 (Li and Godzik 2006) using a clustering
threshold of 0.99 and a word size of 5.
To identify homolog and ortholog sequences, we used the tree-
based identication pipeline described in Yang and Smith (2014)
across all 12 species. All peptide sequences were clustered using an
all- by- all BLASTp search performed using DIAMOND v0.8.29.91
(–evalue 1e- 6 –outfmt 6; Buchnk etal., 2015). e BLASTp results
were then clustered with MCL v14- 137 (Enright etal., 2002) using an
ination value of 1.4. Each of these clusters was then aligned using
MAFFT v7.271 (–genafpair –maxiterate 1000; Katoh and Standley,
2013) and the resulting alignments were trimmed using Phyutility
v2.7.1 (Smith and Dunn, 2008) with minimal column occupancy of
0.1. Codon alignments were produced for each peptide cluster using
PAL2NAL v14 (Suyama etal., 2006) and the nonredundant CDS se-
quences derived above. Tree inference was then performed for each
trimmed alignment using FastTree v2.1.8 (Price etal., 2010) using the
general time- reversible (GTR) model. Because these trees may con-
tain long branches resulting from misassembly, paralogy, or recom-
bination, we trimmed branches that were more than 10 times longer
than its sister or longer than 0.2 substitutions per site. Final homolog
group alignments were created using MAFFT and used for tree in-
ference using RAxML v8.2.9 (Stamatakis, 2014) and the GTRCAT
approximation of the GTR model. Ortholog groups were inferred
using the monophyletic outgroups (MO) method of Yang and Smith
(2014) with G. cuneifolia as outgroup. is method looks for clusters
with monophyletic outgroups, roots the tree, and infers gene duplica-
tion events to identify the subtree with the highest taxa membership
(Yang and Smith, 2014). We ltered orthologous groups to contain
at least 12 taxa and an alignment length of at least 300 bp. Boostrap
branch support in the ortholog trees was assessed using 200 rapid
bootstrap replicates in RAxML. e resulting ortholog alignments,
trees, and bootstrap replicates were used in downstream analyses.
Genomewide dierences between the species
Read mapping and SNP detection—We used BWA v0.7.12 (Li and
Durbin, 2009) to align trimmed Illumina reads from each of the 12
species to the ortholog sequences we created (as described above)
with the default parameters. All samples from both the Bud and
D stages derived from a species were separately aligned to the or-
tholog sequences with high sensitivity. We then sorted and com-
bined alignments from the dierent samples using SAMTOOLS
v1.3.1 (Li etal., 2009) to call variable positions in each species. Only
paired aligned reads were used for SNP calling, using the combined
alignments from all samples in each species. e genotype likeli-
hoods for each individual site were calculated, and allele frequen-
cies were estimated. e ‘MPILEUP’ command in SAMTOOLS was
used to identify SNPs with the parameters ‘- q 30 –C 50 –t SD –t DP
–Q 30 –m 2 –F 0.002 –uf. Genotypes with Phred- scaled genotype
likelihoods below 30 were treated as missing, corresponding to a
genotyping accuracy of at least 99.9%. To reduce the false positive
discovery rate, sites with depth (DP) <30 were also considered to be
missing. VCFtools was used to lter false positive SNPs or paralogs
in each species by excluding sites with depth that was 3× greater
than the mean depth of that species (Danecek etal., 2011).
6 American Journal of Botany
Diversity and divergence—We used the consensus sequence align-
ments from orthology inference (without heterozygous positions)
to calculate sequence diversity, dXY, for all pairwise comparisons of
Achimenes, Eucodonia, and Gesneria. dXY is dened as the number
of diverged sites between two species divided by the alignment
length. We then calculated the relative node depth (RND) of taxa
pairs compared to an outgroup (Feder et al., 2005; Rosenzweig
etal., 2016). e RND is calculated as the divergence between two
species divided by the average divergence between each species and
an outgroup (G. cuneifolia).
Species tree estimations
An initial estimate of phylogenetic relationships was performed
using all 1306 ortholog groups aligned individually and concate-
nated for a maximum likelihood analysis in RAxML under the
GTRGAMMA model. We used 200 rapid bootstrap replicates com-
bined with a tree search from every h bootstrap tree (option –f
a) to assess clade support. Separate runs were performed using a
no partitioning scheme or a scheme partitioning the alignment
with loci treated as separate partitions (n = 1306). No dierence in
topology or bootstrap support was found under dierent partition-
ing schemes.
Under scenarios of high incomplete lineage sorting (ILS), con-
catenation has been demonstrated to have lower power to recon-
struct phylogenetic relationships (Mirarab etal., 2014; Chou etal.,
2015). We addressed the possibility of lineage sorting using two
recently developed methods of coalescent phylogenetic inference:
ASTRAL v4.10.12 (Mirarab and Warnow, 2015) and ASTRID v1.4
(Vachaspati and Warnow, 2015). Both coalescent methods take
advantage of unrooted gene trees to infer species trees under the
coalescent and in the presence of ILS. ASTRAL seeks to nd the
tree that maximizes the number of induced quartets in a set of gene
trees that are shared by the species tree (Mirarab and Warnow,
2015), while ASTRID is an ILS- aware distance- based method that
uses the BIONJ algorithm (Gascuel, 1997; Vachaspati and Warnow,
2015). We also performed 200 multilocus bootstrapping replicates
in both ASTRAL and ASTRID to assess branch support values.
Additionally, the quartet scores were calculated for the ASTRAL
species tree to measure the amount of gene tree conict around
each branch. e quartet scores provide the percentage of quartets
in the gene trees that agree with each branch.
Analyses of introgression
Phylogenetic inference of introgression—Gene ow among
Achimenes can be inferred using a phylogenetic network that allows
for reticulation events (Huson etal., 2011). We created a phyloge-
netic network using PhyloNetworks v0.5.1 (Solís- Lemus and Ané,
2016; Solís- Lemus et al., 2017) and all 1306 gene trees that con-
tained all 12 samples with G. cuneifiolia as the outgroup. We used
SNaQ (Species Networks applying Quartets; Solís- Lemus and Ané,
2016) to evaluate the most likely network (given the species tree and
gene trees) and to calculate γ, the vector of inheritance probabilities
describing the proportion of genes inherited by a hybrid node from
one of its parents. We performed nested analyses that allowed for 0,
1, 2, 3, 4, or 5 hybridization (h) events and compared the negative
log pseudolikelihood score. Optimization in each nested analysis
was performed for 10 independent runs. e run with the lowest
negative log pseudolikelihood score was kept as the best estimate.
A sharp improvement in score is expected until h reaches the best
value and a slower, linear improvement thereaer.
D- statistic—Incongruence between Achimenes species relation-
ships was tested using the D-statistic, also known as the ABBA-
BABA test, which compares counts of discordant site patterns
(Durand etal., 2011). For these analyses, we used ortholog align-
ments that contained all 12 samples that were at least 300 bp long
and removed all gaps. We had a total of 1306 orthologs consisting of
407,343 aligned base pairs and 48,321 variable sites.
e R package HYBRIDCHECK (Ward and Oosterhout, 2015)
was used to count the number of ABBA and BABA site patterns in
four- population phylogenies. e D-statistic was calculated across
12 four- population phylogenies where quartet scores from ASTRAL
or SNaQ indicated admixture or hybridization events. We expect
equal counts of the two site patterns (ABBA and BABA) when in-
complete lineage sorting (ILS) causes discordance. On the other
hand, if discordance is caused by gene ow, we expect the ABBA
site patterns to be more prevalent than the other (i.e., D values will
be positive). Dierences in discordant site pattern counts were
tested using the D-statistic as implemented in HYBRIDCHECK.
Although a full genome alignment (or other linkage information) is
currently not available for Achimenes, our sampled loci likely repre-
sent a random sample of mostly unlinked markers from across the
genome. Under these circumstances, a jackknife approach was used
to test for genomewide variation in incongruence (Meyer et al.,
2012; Eaton and Ree, 2013).
Testing for recent gene ow—As discussed below in the Results
section, we found evidence of gene ow between two pairs of spe-
cies and conicting signal of gene ow between A. admirabilis and
A. erecta (Lam.) H.P. Fuchs. We hypothesized this inconsistency is
due to recent gene ow in sympatry, such as between A. candida
Lindl. and A. misera Lindl., and between A. grandiflora and A. pat-
ens ‘Major, or patterns of incomplete lineage sorting between A.
admirabilis and A. erecta. Recent work has estimated each of these
species originated less than 7 Mya (Roalson and Roberts, 2016). To
test these hypotheses, we investigated patterns of relative divergence
in genes for which A. admirabilis was sister to A. erecta, A. candida
sister to A. misera, and A. grandiflora sister to A. patens ‘Major’.
is analysis is based on two assumptions. First, more recent gene
ow likely results in geographic variation in introgression because,
compared to ancient polymorphism and ILS, there is less time for
novel alleles to spread across populations. Second, loci that were
recently exchanged between populations will show higher sequence
similarity than loci undergoing ILS because introgressed loci have
less time to diverge in each lineage. erefore, we predict that se-
quences from sympatric samples will have more discordant gene
tree topologies as sister taxa, and those sequences showing discord-
ance only in sympatry will have lower between- species divergence
than genes showing the same pattern of discordance in both allo-
patric and sympatric samples.
Within each of the three pairs of species with evidence of gene
ow, there were many gene trees showing each species as sister
to the other. Both ILS and introgression can result in the same
gene tree topology when D values are positive (excess of ABBA
site patterns), but relative divergence between sequences in dier-
ent taxa showing introgression is predicted to be much less than
divergence of sequences that underwent ILS. We predict that if the
excess of genes showing discordant gene trees in each sister pair
2018, Volume 105 Roberts and Roalson—Phylogenomics in Achimenes7
compared to their relationships in the species tree is due to gene
ow, these genes will have lower relative divergence. If the gene
tree discordance in each sister pair is due to ILS, we predict these
genes will not have lower relative divergence compared to the
species tree. To test this prediction, we used the set of orthologs
containing all 12 samples (a total of 1306 genes) to create six sets
of gene trees, two for each sister species comparison. e rst set
of gene trees contained all trees where species A and species B
were sister. e second set of gene trees contained all trees from
the rst set, along with all gene trees that showed the alternate
topology found in the species tree or the SNaQ tree (in the case of
A. grandiflora and A. patens ‘Major’). e RND was calculated for
both sets of gene trees for each species pair and signicance was
assessed using a Student’s t- test. Because our gene tree categories
contained dierent numbers of genes, in addition to testing for
RND dierences between categories, we calculated 95% bootstrap
condence intervals.
Transcriptome assembly and orthology inference
Over 255 million reads and more than 51 Gb were sequenced
from the libraries constructed for this study (Appendix S1; see
Supplemental Data with this article). Libraries that were se-
quenced previously (Roberts and Roalson 2017) for the Bud and
D stages in four species (A. cettoana, A. erecta, A. misera, and
A. patens ‘Major’) were also used in the current study. We car-
ried out de novo transcriptome assembly for the 12 species and
identied orthologs in these assemblies to study the evolution-
ary history of 10 species of Achimenes. Our assembled transcrip-
tomes contained between 58,000 and 111,000 putative transcripts
with a mean N50 of 1698 (Appendix S1). Open reading frames
were detected in 61–74% of the transcripts, resulting in between
21,000 and 24,000 putative genes with 2.5 ± 0.50 putative iso-
forms (Table1). For orthology inference and phylogenetic recon-
struction, we used CDS and excluded noncoding transcripts to
minimize the amount of missing sequence data resulting from
RNA degradation or sequencing errors. Using the monophyletic
outgroups (MO) orthology inference (Yang and Smith, 2014),
we identied 1306 ortholog clusters containing all 12 samples.
Together, these 1306 ortholog clusters contain nearly 2 million
aligned sites and an overall matrix alignment occupancy greater
than 87% (Appendix S2).
Diversity and divergence
Pairwise estimates of sequence similarity (dXY) for the ingroup
ranged from 0.0301 ± 0.0016 (A. longiflora vs. A. pedunculata
Benth.; Appendix S3) to 0.0090 ± 0.0001 (A. grandiflora vs. A. pat-
ens ‘Major’; Appendix S3). e average pairwise sequence similar-
ity across all ingroup species was 0.0212 ± 0.0013 (Appendix S3).
Estimates of pairwise divergence using the RND statistic ranged
from 0.9406 ± 0.0381 (A. longiflora vs. A. pedunculata; Appendix
S3) to 0.3268 ± 0.0223 (A. grandiflora vs. A. patens ‘Major’;
Appendix S3). e average pairwise divergence across all ingroup
species was 0.7263 ± 0.0324 (Appendix S3).
Phylogenetic relationships among Achimenes species
We inferred the phylogenetic relationships among the sampled
Achimenes species using maximum likelihood and two coalescent-
based methods. With 10 of 26 Achimenes species sampled, all
our analyses strongly supported a monophyletic Achimenes with
Eucodonia as sister (BS = 100/100/100; Fig.2C). Achimenes misera
is strongly supported as sister to the species in Clade 1 in both ML
and ASTRAL analyses, which includes A. erecta and A. cettoana,
with much lower support in the ASTRID analysis (BS = 100/100/53;
Fig.2C). In Clade 1, A. cettoana and A. longiflora are strongly sup-
ported as sister in all analyses (BS = 100/100/100; Fig.2C), while A.
admirabilis and A. erecta had strong support in ML and ASTRAL
and much lower support in ASTRID (BS = 100/97/57; Fig. 2C).
In Clade 2, A. candida was separated from the other members of
the clade (Fig.2C). e branch leading to A. antirrhina (DC.) C.V.
Morton, A. pedunculata, A. grandiflora, and A. patens ‘Major’ was
strongly supported in all analyses (BS = 100/100/100; Fig.2C), while
the branch separating A. antirrhina from the other three was less
supported in both coalescent analyses (BS = 100/96/58; Fig.2C).
Examining the quartet scores for each branch produced from
ASTRAL found nearly all with high scores (>60) for the species tree
topology, and two branches that showed much lower scores (<50)
(Appendix S4). e quartet score is proportional to the percentage
of induced quartet trees found in the species tree. Higher quartet
scores indicate a larger proportion of the gene trees that share the
same topology as the inferred species tree. e rst branch with a
low quartet score unites A. admirabilis and A. erecta and quartet
scores for the species tree topology, and the rst alternate shows that
nearly equal proportion of induced quartets support either topol-
ogy (node 5; Appendix S5). e second branch with low quartet
score separates A. antirrhina from A. pedunculata, A. grandiflora,
and A. patens ‘Major’ (node 2; Appendix S5). e quartet scores
TABLE1. Summary of Achimenes CDS transcriptome assemblies.
Species Number genes Number transcripts Mean length N50 length Assembled bases Number of SNPs
Achimenes admirabilis 24,510 73,023 1077 1401 78,615,048 13,252
Achimenes antirrhina 21,610 54,684 993 759 54,299,127 75,207
Achimenes candida 22,051 60,605 960 1209 58,181,853 93,710
Achimenes cettoana 23,278 41,426 1018 1320 42,154,371 11,121
Achimenes erecta 21,583 43,221 926 1167 40,041,570 86,406
Achimenes grandiflora 21,895 48,833 1058 1368 51,677,433 25,097
Achimenes longiflora 21,127 66,491 1052 1356 69,919,131 45,167
Achimenes misera 21,619 47,464 943 1194 44,749,797 84,758
Achimenes patens ‘Major’ 22,110 47,729 918 1149 43,798,476 90,377
Achimenes pedunculata 21,755 69,325 1024 1311 71,019,474 44,551
Eucodonia verticillata 21,436 48,695 1066 1380 51,907,482 36,917
Gesneria cuneifolia 22,169 68,351 957 1209 65,395,389 97,633
8 American Journal of Botany
for the primary topology, rst alternate, and second alternate show
proportions at this branch varied between 25 and 40 (Appendix
S5). Additionally, the node separating Clade 1 and Clade 2 (node 9;
Appendix S5) had quartet scores very close to 50.
Evidence for gene ow among several species pairs
Phylogenetic inference using networks—We used a recently de-
veloped method (Solís- Lemus and Ané, 2016) to infer a phyloge-
netic network of the Achimenes samples from individual gene trees
(Fig.3). Unlike in the ABBA- BABA tests (described below), we used
all 12 samples for this analysis. e inferred phylogenetic network
that best ts our data included three hybrid branches with vectors
of inheritance probabilities (γ) estimated for each (Solís- Lemus and
Ané, 2016; Fig.3; Appendix S6). One hybrid branch led from A.
misera to A. candida (γ = 0.24), another led from A. admirabilis to
A. erecta (γ = 0.47), and a third led from A. grandiflora to A. patens
‘Major’ (γ = 0.40; Fig.3). e hybrid branches leading from A. mi-
sera and A. grandiflora additionally had high bootstrap support of
100 (Fig.3), while the hybrid branch connecting A. admirabilis and
A. erecta had very low boostrap support of 46 (Fig.3). Additionally,
the placement of A. patens ‘Major’ diers in this analysis (Fig.3)
than in the phylogenetic analyses above (Fig.2C). Here, A. patens
‘Major’ is placed sister to A. antirrhina (Fig. 3) with moderately
strong bootstrap support (Fig.3).
D- statistic—We explicitly tested for asymmetry in discordance
patterns using the D-statistic (Green et al., 2010; Durand et al.,
2011). Because this test required sets of four populations, we con-
ducted several tests using dierent combinations of taxa and clades
(Table2). We observed several highly signicant positive D values
for analyses that compared populations both within and between
Clades 1 and 2. Within Clade 1, we observed a signicant posi-
tive D value between A. admirabilis and A. erecta (D = 0.82, P <
0.001) and nonsignicant D values between A. cettoana and either
A. admirabilis or A. erecta. Within Clade 2, signicant positive D
values were observed between A. grandiflora and A. patens ‘Major’
(D = 0.72, P < 0.001) and between A. antirrhina and A. patens
‘Major’ (D = 0.55, P < 0.001), while nonsignicant D values were
calculated between A. antirrhina and either A. pedunculata or A.
grandiflora. Comparing populations between Clade 1 and Clade 2,
we also observed a signicant positive D value between A. candida
and A. misera (D = 0.85, P < 0.001). It is important to note that the
D- statistic is useful to suggest the presence of population admixture
but cannot be used to determine absolute rates of gene ow. e
D-statistic oen cannot distinguish site pattern discordance that
is due to ancient polymorphism/ILS or introgression (Feder etal.,
2005; Goulet etal., 2017). erefore, we applied additional analyses
to determine the most likely gene ow events between Achimenes
populations and to distinguish ILS from gene ow.
Using levels of divergence to test for gene ow—e results from
both D- statistics and the phylogenetic network analyses indicate
admixture between three species pairs: A. admirabilis and A.
erecta, A. candida and A. misera, and A. grandiflora and A. patens
‘Major. First, we investigated the divergence in genes with topol-
ogies showing each species pair as sister with the topologies of
each species pair in the species tree. is method compares the
genetic divergence between genes with discordant topologies to
distinguish between ILS and introgression (Fig.4). We predict that
recently introgressed loci will have discordant tree topologies and
display low interspecic divergence. We found that genes that have
A. candida sister to A. misera had a signicantly lower RND than
genes where each is sister to their respective clades (t = 8.682, df
= 801, P < 0.001; Fig.4A). We also found that genes that have A.
grandiflora sister to A. patens have signicantly lower RND than
genes where A. antirrhina is sister to A. patens (t = 4.727, df =
1404, P < 0.001; Fig.4B). Genes showing A. admirabilis sister to
A. erecta have RND values nearly indistinguishable from genes
showing the alternate topology where A. admirabilis is sister to A.
cettoana and A. longiflora (t = 2.306, df = 527, P = 0.022; Fig.4C).
ese ndings suggest A. candida and A. patens experienced re-
cent introgression from A. misera and A. grandiflora, respectively.
ese ndings additionally suggest some level of ILS between A.
admirabilis and A. erecta causing gene tree discordance between
these populations.
FIGURE 3. Introgression model of Achimenes estimated with SNaQ.
From 1306 genes, h = 3, and rooted with Gesneria cuneifolia. Black
branches: major tree (including hybrid branches with γ > 0.5). Colored
arrows: minor hybrid branches annotated by γ, the vector of heritance
probabilities estimated with SNaQ. Black numbers: bootstrap support for
branches in the major tree, if dierent from 100. Color numbers: boot-
strap support for the placement of minor hybrid branches.
Clade 2Clade 1
γ = 0.24
γ = 0.47
γ = 0.40
TABLE2. Summary of ABBA- BABA tests for population admixture in Achimenes.
P1 P2 P3 O ABBA BABA D P- value
AA AE AC, AL AM 89 895 −0.82 < 0.001
AA AE AC AL 214 136 0.22 1.00
AC, AL AA AE AM 895 89 0.82 < 0.001
AD Clade 2 AM Clade 1 89 998 −0.85 < 0.001
AG AN AT AD, AP 345 178 0.32 0.98
AL AC AA AE 136 214 −0.22 1.00
AN AG AT AD, AP 178 345 −0.32 0.98
AP AG Clade 2 Clade 1 284 215 0.14 0.02
AP AT Clade 2 Clade 1 240 586 −0.42 < 0.001
Clade 2 AD AM Clade 1 998 82 0.85 < 0.001
Clade 2 AP AG Clade 1 1334 284 0.65 < 0.001
Clade 2 AP AT Clade 1 823 240 0.55 < 0.001
Note: AA, A. admirabilis; AC, A. cettoana; AD, A. candida; AE, A. erecta; AG, A. grandiflora; AL, A.
longiflora; AM, A. misera; AN, A. pedunculata; AP, A. patens ‘Major ’; AT, A. antirrhina; Clade 1,
AA, AE, AC, AL; Clade 2, AD, AG, AN, AP, AT; P1, population 1; P2, population 2; P3, population
3; O, outgroup population.
2018, Volume 105 Roberts and Roalson—Phylogenomics in Achimenes9
Hybridization and gene ow are frequent
evolutionary forces that inuence the process
of speciation. New genomic tools provide
an exciting opportunity to test hypotheses
on the eect of gene ow during lineage di-
versication (Gompert and Buerkle, 2016;
Payseur and Rieseberg, 2016; Vallejo- Marín
and Hiscock, 2016). Here we take advantage
of transcriptome sequencing and demon-
strate its utility to reassess phylogenetic re-
lationships in Achimenes and to ask whether
gene ow occurred during the evolution of
this lineage. Using multiple analyses of phy-
logenetic discordance, we show that gene
ow occurred between two pairs of sympa-
tric sister species, and provide evidence that
questions the species status of A. admirabilis.
Phylogenetic relationships among
Achimenes species
Previous phylogenetic hypotheses (Roalson
etal., 2003; Roalson and Roberts, 2016) do
not agree with the phylogenetic hypothesis
presented here. ese studies have indi-
cated moderate to strong support for species
placement in Achimenes within three dis-
tinct clades (Clade 1, Clade 2, and Clade 3;
sensu Roalson etal., 2003). Resolution was
lacking, however, at the crown of the genus
to indicate how these three clades were re-
lated to one another (Roalson etal., 2003;
Roalson et al., 2005; Roalson and Roberts, 2016). e previous
phylogenetic hypothesis of Roalson etal. (2003) was generated
using two loci (nuclear ITS and plastid trnL-F spacer), while the
hypothesis presented here uses >1300 loci and demonstrates the
utility of transcriptome- based phylogenomic approaches to reas-
sess previous Sanger- based hypotheses. e level of discordance
between the phylogenetic hypothesis of this study and Roalson
etal. (2003) could be due to gene sampling eects. e two loci
used by Roalson etal. (2003) likely belong to gene families with
alternate histories from the species tree presented here. Our re-
sults suggest two distinct clades in Achimenes, Clade 1 and Clade
2 (Fig.2C), with strong bootstrap support from all four methods
employed (BS = 100/100/100/100; Fig.2C). While these estimates
provide strong bootstrap support, the quartet score for the pri-
mary topology at the node separating Clade 1 and Clade 2 was
close to 50 (node 9; Appendix S4). While not indicative of strong
ILS at this node, the decreased quartet score could reect lower
support for the placement of A. misera in Clade 1 (Fig.2C). An
increased sampling of species, particularly some that were placed
in the original Clade 3 of Roalson etal. (2003), might shed more
light on this apparent discordance. Lastly, the addition of sam-
ples from Smithiantha, a small herbaceous genus closely related
to Eucodonia and Achimenes, and Solenophora, a genus of woody
shrubs thought to be closely allied to Achimenes, would also pro-
vide additional data to test the apparent monophyly of Achimenes
presented here.
Results also indicate that A. patens ‘Major’ is strongly sup-
ported as sister to A. grandiflora in all phylogenetic analyses (BS =
100/100/100; Fig.2C). is result adds support to our hypothesis
that A. patens ‘Major’ experienced some level of gene ow with A.
grandiflora, particularly given the numerous morphological sim-
ilarities between them. Given the abundance of discordant gene
trees, it is not surprising that previous analyses based on limited
gene sampling suggested dierent relationships. Additional dis-
cordant patterns of topology and branch support were found, par-
ticularly when comparing species tree methods. All relationships
were strongly supported with BS = 100 when using ML (Fig.2C),
while two branches were slightly less well supported using ASTRAL
(BS ≥ 96; Fig.2C), and three branches had weak support when us-
ing ASTRID (BS ≤ 58; Fig.2C). e branch showing a relationship
between A. admirabilis and A. erecta and the branch separating A.
antirrhina from A. grandiflora, A. patens ‘Major’, and A. pedun-
culata, were both the same branches where quartet support was
low (Appendix S4), indicating ILS or admixture at these branches
(Fig.2C). Both ASTRAL and ASTRID, using dierent approaches
to estimate a species tree, have been shown in some data sets to re-
duce branch support when there is high variance in gene tree topol-
ogies (Esselstyn etal., 2017).
Additionally, SNaQ analysis showed A. erecta sister to the rest of
Clade 1, minus A. misera (BS = 57; Fig.3), rather than sister to A.
admirabilis (BS = 100/97/57; Fig.2C). Furthermore, the relation-
ship of A. misera to Clade 1 was strongly supported in the SNaQ
FIGURE4. Tests of recent introgression. (A) Interspecic divergence for genes that show A. can-
dida as sister to A. misera (‘AD, AM’) and as sister to both A. misera and Clade 2 (‘AD, AM both’). (B)
Interspecic divergence for genes that show A. patens ‘Major’ as sister to A. grandiora (‘AG, AP’)
and as sister to both A. grandiora and A. antirrhina (‘AG, AP both’). (C) Interspecic divergence for
genes that show A. admirabilis as sister to A. erecta (‘AA, AE’) and as sister to both A. erecta and A.
cettoana/A. longiora (‘AA, AE both’). Results from t- tests comparing the means are shown. Bars
indicate 95% condence intervals as calculated using bootstrap resampling.
10 American Journal of Botany
analysis (BS = 100; Fig.3), but weakly supported in the ASTRID
analysis (BS = 53; Fig.2C). Many coalescent- based methods (such
as ASTRAL and ASTRID) work under the assumption that ILS is
the only source of gene tree discordance (Mirarab and Warnow,
2015; Vachaspati and Warnow, 2015), while ignoring the presence
of gene ow. ese species tree methods are not robust to such vio-
lations even with large numbers of well- constructed trees and may
be inconsistent under gene ow (Solís- Lemus etal., 2016). SNaQ
allows for both ILS and gene ow (Solís- Lemus and Ané, 2016). e
branches showing very weak support in ASTRID were also branches
on the tree where other analyses indicated ILS or gene ow. e
distance- based algorithm of ASTRID may be less robust under high
gene ow than the quartet- based algorithm of ASTRAL (Davidson
etal., 2015). Our results demonstrate that estimating evolutionary
relationships in lineages where both ILS and gene ow occur (such
as Achimenes) remains a challenging endeavor and advocates for
more extensive use of network- based approaches that can account
simultaneously for both processes (Solís- Lemus and Ané, 2016).
Eects of gene ow during speciation
Phylogenetic network analysis indicated the evolutionary history of
these species includes some level of gene ow (Fig.3). Additionally,
four- taxon analyses of population admixture provided additional
evidence for gene ow between two Achimenes species pairs, from
A. misera to A. candida, and from A. grandiflora to A. patens
‘Major’ (Table2). While the D-statistic can detect, but not quantify,
introgression or admixture, many studies have shown it to be robust
when used on a genome- wide scale (Green etal., 2010; Eaton and
Ree, 2013). However, it can also be stochastic when applied over
small windows and is sensitive to within- species diversity (Martin
et al., 2015), sometimes producing conicting results. erefore,
we caution the over interpretation of its meaning outside of pro-
viding an indication of potential admixture that should be more
extensively evaluated using alternative measures, such as the
tistic (Martin etal., 2015). Using exemplars for this study provides
initial evidence that gene ow occurred within these species pairs
sometime in the past and provides a starting point for additional
sampling of individuals to explore the prevalence of gene ow and
hybridization within sympatric populations.
Combining the results of each analysis, we found two consist-
ent sister pairs that exhibited signs of gene ow: A. candida and A.
misera, and A. grandiflora and A. patens ‘Major. Each pair consists
of species that display remarkable similarity in oral form (Fig.1).
Gene ow among species that share similar pollinators has been
observed in other plant groups, including many orchid lineages
(Cortis etal., 2009; Gögler etal., 2015). Our estimates of gene ow
from A. candida to A. misera and from A. grandiflora to A. patens
‘Major’ are too low for an early hybrid, and are comparable to levels
found in other species with moderate levels of hybridization (Cahill
etal., 2016; Solís- Lemus and Ané, 2016). Together with the similar-
ities in oral form, these results indicate gene ow between the two
pairs is not unexpected and might suggest that gene ow occurred
through visitations by a common pollinator.
Hybrids between A. grandiflora and A. patens are observed in
the eld (Wiehler, 1983; Ramírez Roa, 1987) and have high fer-
tility when crossed in the greenhouse (Cooke and Lee, 1966).
Experimental crosses between the two species indicate that these
hybridization events produce stainable pollen at rates of 88%
(Cooke and Lee, 1966). Achimenes grandiflora is among the most
widespread species in the genus and is sympatric with numer-
ous species (Fig. 2B; Wiehler, 1983; Ramírez Roa, 1987). Before
A. patens was formally described as a species in 1840, it was con-
sidered synonymous with A. grandiflora because of the striking
similarity of their owers and their overlapping geographic dis-
tributions (Gordon, 1846). While these similarities in oral form
are certainly an example of oral convergence, there are some dif-
ferences. Particularly, the corolla spur in A. patens can be pointed
and elongated, whereas, the corolla spur in A. grandiflora is blunt
and short. Our sample of A. patensMajor’ showed intermediate
characters, particularly A. patens ‘Major’ has a similar short, blunt
spur to A. grandiflora, while retaining other oral and vegetative
characteristics of A. patens. We initially hypothesized that A. pat-
ens ‘Major’ may have experienced some introgression in the past,
likely from A. grandiflora. Both A. grandiflora and A. patens ‘Major’
share the closest genetic similarity among all pairwise comparisons
(Appendix S3) and were found sister in all species tree reconstruc-
tions (Fig.2C), contrary to previous analyses (Roalson etal., 2003).
Our results from all analyses of gene ow indicate that introgression
has likely occurred between these species at some point in the past.
Achimenes misera has been considered a reproductively isolated
species (Cooke and Lee, 1966; Ramírez Roa, 1987). While nearly all
of Achimenes are diploids (n = 11; Cooke and Lee, 1966; Ramírez
Roa, 1987), A. misera (and A. erecta) is a polyploid (n = 22; Cooke
and Lee, 1966; Ramírez Roa, 1987). It was thought that this dierence
in chromosome number might be the limiting factor in the forma-
tion of hybrids (Ramírez Roa, 1987). e suggestion by our analyses
that gene ow occurred from A. misera to A. candida was therefore
surprising given these reports. Achimenes candida was previously
shown to have the ability to produce stainable hybrid pollen with
many other species, although only with species belonging to Clade
2 (Fig. 2B; Cooke and Lee, 1966; Wiehler, 1983), which does not
include A. misera. When we considered the putative paralog contri-
butions of each species to the others’ genome, we found 323 putative
paralogs of A. candida and A. misera (25% of 1306 orthologs) in our
data set. Within that group of paralogs, 62 (19% of the paralogs or 5%
of the total) were contributed from A. candida to A. misera and 184
(57% of the paralogs or 14% of the total) were contributed from A.
misera to A. candida. ese numbers would suggest that introgres-
sion occurred between these species sometime in the recent past. In
order for ~5% of the A. candida genome to have introgressed from
A. misera, a possible route may have been through an initial hybrid-
ization event with A. misera (both diploids at this point) followed
by several backcrossing events to A. candida (Fig.5A). Similarly for
~14% of the A. misera genome to be introgressed from A. candida,
one possible route would be through an initial hybridization event
with A. candida (both diploids) followed by several backcrossing
events to A. misera, eventually ending with an autopolyploidy event
that creates the polyploid A. misera (Fig.5B). ese two scenarios t
extremely well with the data and oer testable hypotheses for future
studies. More extensive population sampling of these two species in
Mexico, particularly in areas where they are sympatric, would pro-
vide further insight into how extensive introgression and gene ow
contributed to patterns of diversity.
Given that A. candida and A. misera have rather small, unas-
suming owers, the kind typically of little interest to horticultural
hybridizers of Achimenes, we did not initially hypothesize that gene
ow or hybridization would have occurred during the history of
this pair. Phylogenetic network analyses found a strongly supported
hybrid branch between these species (BS = 100; Fig.3), while the
2018, Volume 105 Roberts and Roalson—Phylogenomics in Achimenes11
genetic similarity between the two species was also moderate com-
pared to other pairwise comparisons (Appendix S3). Achimenes mi-
sera has also received interest for its close anity to another species,
A. warszewicziana (Regel) H.E. Moore. is species also displays
similar oral form to both A. candida and A. misera, is pollinated
by bees, and is found in sympatric locations with both A. candida
and A. misera. Until very recently, Achimenes warszewicziana was
considered to be synonymous with A. misera. It is now regarded as
a distinct species. Sampling of these three species may reveal more
extensive patterns of gene ow and introgression between sympa-
tric populations that have similar owers and similar pollinators.
Gene ow among sympatric species that share pollinators has
been studied for decades in many dierent plant systems (Beattie,
1976; Campbell, 1985; Soliva and Widmer, 2003; Gögler etal., 2015).
In gesneriads, the contributions of pre- and post- mating barriers to
gene ow in the maintenance of species barriers has not been exten-
sively studied, outside of sympatric Hawaiian Cyrtandra (Johnson
etal., 2015). Our results oer preliminary genomic evidence of ex-
tensive gene ow within Achimenes, and provide a starting point
to address other patterns of sympatric gene ow within other
Neotropical gesneriads such as Sinningia (Perret etal., 2007). Pre-
mating reproductive isolation between sympatric species is oen
associated with dierences in owering time (Soliva and Widmer,
1999; Savolainen etal., 2006), but changes in ower morphology
also contribute to reproductive isolation through specialization to
dierent pollinators (Grant, 1971; Ramsey etal., 2003; Fenster etal.,
2004), or by limiting pollen transfer between species with similar
pollinators (Armbruster etal., 1994; Wolf etal., 2001). Hybridization
experiments in Neotropical gesneriads reveal high pollen stainabil-
ity and fertility among interspecic hybrids and indicate that the
FIGURE5. Scenarios of introgression between A. candida and A. misera. (A) An introgression model for A. candida begins with a hybridization event
with A. misera and is followed by several backcrossing events to A. candida. This scenario leads to an estimated 6.25% of the A. candida genome
being shared with A. misera, closely matching the 5% estimated from the transcriptome data. (B) An introgression model for A. misera begins with a
hybridization event with A. candida, followed by several backcrossing events to A. misera, and ending with a recent autopolyploidy event that creates
the polyploid A. misera. This scenario leads to an estimated 12.5% of the A. misera genome being shared with A. candida, closely matching the 14%
estimated from the transcriptome data. Pie charts indicate the estimated proportion of the genome shared with A. candida or A. misera (colored in
black and white, respectively). The ploidy of each individual is these scenarios is indicated below each circle.
A. candida A. misera
n=11 n=11
n=11 n=11
A. candida A. misera
n=11 n=11
n=11 n=11
tion of
genome shared
A. misera
1/2 (50%)
1/4 (25%)
1/8 (12.5%)
1/16 (6.25%)
misera to A. candida
Under scenario:
Transcriptome data:
62 paralogs, 5%
Proportion of
genome shared
with A. candida
1/2 (50%)
1/4 (25%)
1/8 (12.5%)
A. candida to A. misera
Under scenario:
Transcriptome data:
184 paralogs, 14%
1/8 (12.5%)
12 American Journal of Botany
eective isolating mechanisms between species may not be genetic,
but external physiological, spatial, or ecological barriers (Wiehler,
1983). No studies have looked at whether pre- or post- mating bar-
riers contribute to reproductive isolation among closely related spe-
cies in Achimenes or other Neotropical gesneriads.
Asymmetric gene ow
Our phylogenetic analyses suggest that the evolutionary history of
these lineages involved asymmetric gene ow from A. misera into
A. candida, and from A. grandiflora into A. patens ‘Major’ (Table2;
Fig.3). is would presumably occur because hybrids backcrossed
into A. candida and A. patens ‘Major’ more than with A. misera and
A. grandiflora, respectively. e direction of gene ow between these
two species pairs agrees with our predictions of gene ow based on
morphological similarities and range overlap in central Mexico.
Directional bias in gene ow from one species into another can
be important for determining the direction of evolutionary change
and species succession (Petit etal., 2004). ere has been long-
standing interest in understanding the factors that drive asymmet-
rical gene ow in plants. Some of the underlying factors can include
mating system variation (Lewis and Crowe, 1958), the relative pro-
portions of parent species (Burgess etal., 2005), and dierences in
the tness of reciprocal crosses (Tin etal., 2001). Alternatively,
asymmetric gene ow could reect demographic processes related
to species range expansion, which has been implicated as a major
determinant of gene ow in a wide array of plants and animals
(Currat etal., 2008). Our results indicate asymmetries in gene ow
from two of the most widely distributed species.
Achimenes grandiflora has one of the largest ranges throughout
Mexico and Central America, is sympatric with a high number of
species, and can produce hybrids with the largest number of other
species (Cooke and Lee, 1966; Ramírez Roa, 1987). Achimenes mi-
sera is also widespread throughout Central America, but had previ-
ously been considered reproductively isolated (Ramírez Roa, 1987).
In many natural plant populations, selection remains the primary
mechanism implicated in determining patterns of hybridization
and introgression (Lexer etal., 2005; Whitney etal., 2006). Further
exploration and sampling of populations sympatric with both A.
grandiflora and A. misera will provide data allowing us to determine
how species boundaries are maintained and what role selection and
demographic processes play in the patterns found in Achimenes.
Additional genomic sequencing of more neutral loci than those
used in the current study would further allow more sophisticated
estimates of migration and gene ow (Gronau etal., 2011).
Geographic and ecological patterns of gene ow and
We performed genomic analyses of gene ow using 10 transcrip-
tomes that represent 10 species of Achimenes. ese plants are oen
found in sympatry and allopatry throughout Mexico and Central
America. Species can be found from sea level upward to 3000 m
growing in Quercus and Pinus forests, but also found in transitional
zones between forests and arid subtropical shrubs. Geographic pat-
terns can additionally be found among the pollination syndromes.
ose species that are bee- pollinated (e.g., A. candida and A. mi-
sera) tend to be more narrowly endemic in western and southern
Mexico (Figs.2A, B), while species pollinated by butteries (e.g., A.
grandiflora and A. longiflora) and hummingbirds (e.g., A. antirrhina
and A. erecta) are more widespread throughout Mexico and Central
America (Figs.2A, B). ese emerging patterns contribute to our
understanding of diversication processes in this lineage.
Achimenes candida and A. misera can be found in Chiapas,
Mexico and inhabit similar forest habitats and elevations through-
out their ranges (Figs.2A, B; Ramírez Roa, 1987). ese overlaps
also extend to owering times. Achimenes misera owers from
April through October and A. candida owers from July through
August (Ramírez Roa, 1987). ese ecological and phenological
factors may contribute to gene ow in contact areas where species
may be pollinated by similar bees during similar owering times.
More extensive work will need to be done in Chiapas to test these
hypotheses within sympatric populations.
e specimen of A. patens sampled in this study was A. patens
‘Major, originally described as a “rst- class variety” (Moore, 1859).
e origins of this variety are unclear, but it displays similarity to
both A. grandiflora and A. patens, with the most obvious dierence
being that A. patens ‘Major’ contains a short, blunt corolla spur sim-
ilar to A. grandiflora. Given the clear similarities between A. patens
‘Major’ and A. grandiflora, we hypothesized that A. patens ‘Major’
may represent a case of introgression from A. grandiflora. We en-
vision two potential scenarios for the origins of this variety. First,
this variety could represent a natural hybrid that was brought into
cultivation from Mexico. Rhizomes of A. patens were rst brought
to England by a Mr. Hartweg in 1846 from Zitacuaro, Mexico, a
location where populations of A. grandiflora and A. patens are
found in sympatry (Gordon, 1846). Both species inhabit similar
forest habitats in higher elevations, upwards of 1800 m (Ramírez
Roa, 1987). Second, this variety could represent a horticultural hy-
brid whose origins might be found among hybridizations that took
place during the peak popularity of magic owers in Victorian Era
England. Our results from all analyses provide support for intro-
gression (Table2; Figs.2B and 3) between A. patens ‘Major’ and A.
grandiflora, but without a better record for A. patens ‘Major’ we can
only speculate about its origins.
Achimenes erecta is one of the most widespread species through-
out Central America and some populations can be found in the
Caribbean (Fig.2A; Ramírez Roa, 1987). e original type specimen
of this species was sent to England from Jamaica in 1778 (Fuchs,
1963). Considerable morphological variation is known from across
the range of this species, including vegetative and reproductive
characters (Wiehler, 1983; Ramírez Roa, 1987). Many varieties col-
lected in dierent locations in Mexico, Central America, and the
Caribbean have been brought into cultivation. Before A. admira-
bilis was described as a species (Wiehler, 1992), it was considered a
variety of A. erecta, but having characteristics similar to A. cettoana
(Ramírez Roa, 1987). Leaves in both A. cettoana and A. admirabi-
lis are elliptic- linear and owers are curved and glabrous, while A.
erecta has lanceolate leaves and slightly curved, puberulent owers.
Achimenes admirabilis is only found from a few populations within
Oaxaca and our analyses indicate that the patterns of gene tree dis-
cordance between A. admirabilis and A. erecta we found may partly
be due to ILS, possibly with a low level of gene ow (Figs.2C and 3).
ese results suggest the status of A. admirabilis should be reevalu-
ated with additional sampling of A. erecta and A. admirabilis from
throughout their ranges. e high variation in A. erecta may sug-
gest that other varieties exist. Potential scenarios for the status of
A. admirabilis exist, and given more extensive sampling of A. erecta
may include two possibilities. (1) A. erecta may include multiple
lineages that are as distinct as A. admirabilis, possibly leading to the
2018, Volume 105 Roberts and Roalson—Phylogenomics in Achimenes13
recognition of more species lineages. (2) Achimenes erecta may be a
monophyletic lineage and A. admirabilis is in the process of diver-
gence from A. erecta sensu lato. e evidence for gene ow between
the two samples included here suggests that sampling A. erecta
from across its range and morphological variability, in combination
with A. admirabilis samples, will be necessary to more fully address
the relationship between these two putative lineages.
We took advantage of the sister relationship between A. admira-
bilis and A. erecta to elucidate whether the observed gene tree dis-
cordances were due to ILS or introgression. D- statistics alone oen
cannot distinguish site patterns resulting from ILS and ancient pol-
ymorphism or recent hybridization (Feder etal., 2005; Rosenzweig
etal., 2016; Goulet etal., 2017). erefore, we compared divergence
in genes with our discordant phylogenetic signals (Fig. 4; Roda
etal., 2017). We found nearly equal numbers of genes supporting
A. admirabilis as sister to A. erecta or supporting A. admirabilis
sister to A. cettoana and A. longiflora (Appendices S4 and S5).
Furthermore, we found that those genes showing A. admirabilis sis-
ter to A. erecta were not signicantly less diverged than those gene
trees showing both the discordant topology and the species tree
topology (Fig.4C). ese results might suggest that the status of A.
admirabilis being a distinct species from A. erecta be reconsidered.
As with other genome- wide analyses inferring gene ow, our
study depends on analyzing patterns across many loci in a limited
number of individuals. Sequencing many more individuals and spe-
cies from across the range of Achimenes throughout Mexico and
Central America will provide stronger estimates of the timing and
amount of gene ow across the landscape, particularly for popula-
tions found in allopatry and sympatry with other species.
Our transcriptome analyses provide evidence of gene ow and in-
trogression during the evolution of Achimenes. Multiple phylog-
enomic analyses of gene ow indicate introgression has occurred
between at least two species pairs that share pollinators and are
found in sympatry. ese analyses also call into question the species
status of A. admirabilis. Although the analyses applied here were
originally designed for a small number of samples (Green et al.,
2010; Payseur and Rieseberg, 2016), we acknowledge that some of
our results could benet by increasing the sampling. Particularly,
sampling populations from across the range of Achimenes in
Mexico, including both allopatric and sympatric individuals, will
allow for a better quantication of the timing, direction, and mag-
nitude of gene ow. Transcriptome sequencing approaches provide
extensive genomic resources useful for studies of biodiversity that
allows us to investigate both the patterns and processes involved
in the evolution of tropical lineages. Lastly, the current study high-
lights interesting patterns of gene ow among species of Achimenes
and provides the basis for further phylogenomic and phylogeo-
graphic studies into the evolution and diversication of this colorful
and diverse lineage of gesneriads.
We thank Associate Editor John Freudenstein, Lucy Allison,
Kimberly Hansen, Nan Jiang, Joseph Kleinkopf, and two anon-
ymous reviewers for thoughtful comments on the manuscript;
Joanna Kelley for access to laboratory facilities and helpful dis-
cussion on the manuscript; Corey Quackenbush for valuable in-
sight on library preparation; Michael Ne and Karen Sanquinet
for access to their Real- Time PCR machine; and Chuck Cody for
maintaining the growth and happiness of our gesneriad collection.
e molecular work of this study was conducted in the Kelley and
Roalson Labs in the School of Biological Sciences, Washington State
University. e sequencing work of this study was conducted in the
Genomics Core Lab at Washington State University, Spokane. e
Elvin McDonald Research Endowment Fund from e Gesneriad
Society [to W.R.R], the Global Plant Sciences Initiative Fellowship
[to W.R.R], and a NSF Doctoral Dissertation Improvement Grant
DEB- 1601003 [to W.R.R. and E.H.R.] supported this research.
Raw reads for the 64 sequenced libraries generated in this study
are deposited in the NCBI Sequence Read Archive (BioProject:
PRJNA401042). Raw reads for the 12 sequenced libraries of Roberts
and Roalson (2017) are deposited in the NCBI Sequence Read
Archive (BioProject: PRJNA340450). Assembled sequences, data
les, alignments, and trees are available from the Dryad Digital
Repository: Scripts and data
for analyses are available from
Additional Supporting Information may be found online in the
supporting information tab for this article.
Anderson, E. 1949. Introgressive hybridization. Wiley, New York.
Andrews, S. 2010. FastQC: a quality control tool for high throughput sequence
data. Available at:
fastqc [accessed 12 December 2016].
Armbruster, W. S., M. E. Edwards, and E. M. Debevec. 1994. Floral character
displacement generates assemblage structure of western Australian trigger-
plants (Stylidium). Ecology 75: 315–329.
Beattie, A. J. 1976. Plant dispersion, pollination, and gene ow in Viola.
Oecologia 25: 291–300.
Becker, J. 2008. e Gesneriad Register 2008: Achimenes. e Gesneriad Society,
Bolger, A. M., M. Lohse, and B. Usadel. 2014. Trimmomatic: a exible trimmer
for Illumina sequence data. Bioinformatics 2: 8–13.
Buchnk, B., C. Xie, and D. H. Huson. 2015. Fast and sensitive protein align-
ment using DIAMOND. Nature Methods 12: 59–60.
Burgess, K. S., M. Morgan, L. Deverno, and B. C. Husband. 2005. Asymmetrical
introgression between two Morus species (M. alba, M, rubra) that dier in
abundance. Molecular Ecology 14: 3471–3483.
Cahill, J. A., Z. Fan, I. Gronau, J. Robinson, J. P. Bollinger, B. Shapiro, J. Well,
and R. K. Wayne. 2016. Whole- genome sequence analysis shows that two
endemic species of North American wolf are admixtures of the coyote and
gray wolf. Science Advances 2: e1501714.
Campbell, D. R. 1985. Pollen and gene dispersal: the inuences of competition
for pollination. Evolution 39: 418–431.
Chou, J., A. Gupta, S. Yaduvanshi, R. Davidson, M. Nute, S. Mirarab, and
T. Warnow. 2015. A comparative study of SVDquartets and other
14 American Journal of Botany
coalescent- based species tree estimation methods. BMC Genomics 16(Suppl
10): S2.
Cooke, J. F., and R. E. Lee. 1966. Hybridization within and between Achimenes P.
Br. and Smithiantha Kuntze (Gesneriaceae). Baileya 14: 92–101.
Cortis, P., N. J. Vereecken, F. P. Schiestl, M. R. Barone Lumaga, A. Scrugli, and
S. Cozzolino. 2009. Pollinator convergence and the nature of species’ bound-
aries in sympatric Sardinian Ophrys (Orchidaceae). Annals of Botany 104:
Coyne, J. A., and H. A. Orr. 1997. “Patterns of speciation in Drosophila” revis-
ited. Evolution 51: 295–303.
Currat, M., M. Reudi, R. J. Petit, and L. Excoer. 2008. e hidden side of inva-
sions: massive introgression by local genes. Evolution 62: 1908–1920.
Danecek, P., A. Auton, G. Abecasis, C. A. Albers, E. Banks, M. A. DePristo, R. E.
Handsaker, et al. 2011. e variant call format and VCFtools. Bioinformatics
15: 2156–22158.
Davidson, R., P. Vachaspati, S. Mirarab, and T. Warnow. 2015. Phylogenomic
species tree estimation in the presence of incomplete lineage sorting and
horizontal gene transfer. 2015. BMC Genomics 16 (Suppl 10): S1.
Des Marais, D. L., and M. D. Rausher. 2010. Parallel evolution at multiple levels
in the origin of hummingbird pollinated owers in Ipomoea. Evolution 64:
Durand, E. Y., N. Patterson, D. Reich, and M. Slatkin. 2011. Testing for ancient
admixture between closely related populations. Molecular Biology and
Evolution 28: 2239–2252.
Eaton, D. A. R., and R. H. Ree. 2013. Inferring phylogeny and introgression using
RADseq data: an example from ower plants (Pedicularis: Orobanchaceae).
Systematic Biology 62: 689–706.
Enright, A. J., S. Van Dongen, and C. A. Ouzounis. 2002. An ecient algorithm
for large- scale detection of protein families. Nucleic Acids Research 30:
Esselstyn, J. A., C. H. Oliveros, M. T. Swanson, and B. C. Faircloth. 2017.
Investigating dicult nodes in the placental mammal tree with expanded
taxon sampling and thousands of ultraconserved elements. Genome Biology
and Evolution 9: 2308–2321.
Feder, J. L., X. Xie, J. Rull, S. Velez, A. Forbes, B. Leung, H. Dambroski, et al.
2005. Mayr, Dobzhansky, and Bush and the complexities of sympatric spe-
ciation in Rhagoletis. Proceedings of the National Academy of Sciences, USA
102: 6573–6580.
Fenster, C. B., W. S. Armbruster, P. Wilson, M. R. Dudash, and J. D. omson.
2004. Pollination syndromes and oral specialization. Annual Review of
Ecology, Evolution, and Systematics 35: 375–403.
Folk, R. A., J. R. Mandel, and J. V. Freudenstein. 2017. Ancestral gene ow and
parallel organellar genome capture result in extreme phylogenomic discord
in a lineage of angiosperms. Systematic Biology 66: 320–337.
Fuchs, H. P. 1963. Achimenes erecta. Acta Botanica Neerlandica 12: 15.
Gascuel, O. 1997. BIONJ: an improved version of the NJ algorithm based on a
simple model of sequence data. Molecular Biology and Evolution 14: 685–695.
Gögler, J., J. Stökl, P. Cortis P, H. Beyrle, M.R. Barone Lumaga, S. Cozzolino,
and M. Ayasse. 2015. Increased divergence in oral morphology strongly
reduced gene ow in sympatric sexually deceptive orchids with the same
pollinator. Evolutionary Ecology 29: 703–717.
Gompert, Z., and C. A. Buerkle. 2016. What, if anything, are hybrids: enduring
truths and challenges associated with population structure and gene ow.
Evolutionary Applications 9: 909–923.
Gordon, G. 1846. Some accounts of Achimenes patens, with its cultivation, and
that of the species allied to it. The Journal of the Horticultural Society of
London 1: 257–260.
Goulet, E. B., F. Roda, and R. Hopkins. 2017. Hybridization in plants: old ideas,
new techniques. Plant Physiology 173: 65–78.
Grabherr, M.G., B.J. Haas, M. Yassour M, J.Z. Levin, D.A. ompson, I. Amit,
X. Adiconis, et al. 2011. Trinity: reconstructing a full- length transcriptome
without a genome from RNA- Seq data. Nature Biotechnology 29: 644.
Grant, V. 1971. Plant speciation. Columbia University Press, New York, NY.
Green, R. E., J. Krause, A. W. Biggs, T. Maricic, U. Stenzel, M. Kircher, N.
Patterson, et al. 2010. A dra sequence of the Neandertal genome. Science
328: 710–722.
Gronau, I., M. J. Hubisz, B. Gulko, C. G. Danko, and A. Siepel. 2011. Bayesian in-
ference of ancient human demography from individual genome sequences.
Nature Genetics 43: 1031–1034.
Haas, B. J., A. Papanicolaou, M. Yassour, M. Grabherr, P. D. Blood, J. Bowden,
M. B. Couger, et al. 2013. De novo transcript sequence reconstruction from
RNA- seq using the Trinity platform for reference generation and analysis.
Nature Protocols 8: 1494–1512.
Hanson, K. D., S. E. Brenner, and S. Dudoit. 2010. Biases in Illumina tran-
scriptome sequencing caused by random hexamer priming. Nucleic Acids
Research 36: e105.
Hoballah, M. E., T. Gübitz, J. Stuurman, L. Broger, M. Barone, T. Mandel, A.
Dell’Olivo, et al. 2007. Single gene- mediated shi in pollination attraction in
Petunia. The Plant Cell 19: 779–790.
Huson, D. H., R. Rupp, and C. Scornavacca. 2011. Phylogenetic networks: con-
cepts, algorithms, and applications. Cambridge University Press, Cambridge,
United Kingdom.
Johnson, M. A., D. K. Price, J. P. Price, and E. A. Stacy. 2015. Postzygotic barriers
isolate sympatric species of Cyrtandra (Gesneriaceae) in Hawaiian montane
forest understories. American Journal of Botany 102: 1870–1882.
Joly, S., P. A. McLenachan, and P. J. Lockhart. 2009. A statistical approach for
distinguishing hybridization and incomplete lineage sorting. The American
Naturalist 174: E54–E70.
Katoh, K., and D. M. Standley. 2013. MAFFT multiple sequence alignment
soware version 7: improvements in performance and usability. Molecular
Biology and Evolution 30: 772–780.
Kubatko, L. S. 2009. Identifying hybridization events in the presence of coales-
cence via model selection. Systematic Biology 58: 478–488.
Lewis, D., and L. K. Crowe. 1958. Unilateral interspecic incompatibility in ow-
ering plants. Heredity 12: 233–256.
Lexer, C., M. F. Fay, J. A. Joseph, M. S. Nica, and B. Heinze. 2005. Barrier to gene
ow between two ecologically divergent Populus species, P. alba (white pop-
lar) and P. tremula (European aspen): the role of ecology and life history in
gene introgression. Molecular Ecology 14: 1045–1057.
Li, H., and R. Durbin. 2009. Fast and accurate short read alignment with
Burrows- Wheeler transform. Bioinformatics 25: 1754–1760.
Li, H., B. Handsaker, A. Wysoker, T. Fennell, J. Ruan, N. Homer, G. Marth, et al.
2009. e sequence alignment/map format and SAMtools. Bioinformatics
25: 2078–2079.
Li, W., and A. Godzik. 2006. Cd- hit: a fast program for clustering and comparing
large sets of protein or nucleotide sequences. Bioinformatics 22: 1658–1659.
Linder, C. R., and L. H. Rieseberg. 2004. Reconstructing patterns of reticulate
evolution in plants. American Journal of Botany 91: 1700–1708.
Mallet, J., N. Besansky, and M. W. Hahn. 2016. How reticulated are species?
BioEssays 38: 140–149.
Martén-Rodríguez, S., A. Almarales-Castro, and C. B. Fenster. 2009. Evaluation
of pollination syndromes in Antillean Gesneriaceae: evidence for bat, hum-
mingbird and generalized owers. Journal of Ecology 97: 348–359.
Martén-Rodríguez, S., C. B. Fenster, I. Agnarsson, L. E. Skog, and E. A. Zimmer.
2010. Evolutionary breakdown of pollination specialization in a Caribbean
plant radiation. New Phytologist 188: 403–417.
Martin, S. H., J. W. Davey, and C. D. Jiggins. 2015. Evaluating the use of ABBA-
BABA statistics to locate introgressed loci. Molecular Biology and Evolution
32: 244–257.
McBreen, K., and P. J. Lockhart. 2006. Reconstructing reticulate evolutionary
histories of plants. Trends in Plant Science 11: 398–404.
Meyer, M., M. Kircher, M.-T. Gansauge, H. Li, F. Racimo, S. Mallick, J. G.
Schraiber, et al. 2012. A high- coverage genome sequence from an archaic
Denisovan individual. Science 338: 222–226.
Mirarab, S., M. Shamsuzzoha Bayzid, and T. Warnow. 2014. Evaluating sum-
mary methods for multilocus species tree estimation in the presence of in-
complete lineage sorting. Systematic Biology 65: 366–380.
Mirarab, S., and T. Warnow. 2015. ASTRAL- II: coalescent- based species tree es-
timation with many hundreds of taxa and thousands of genes. Bioinformatics
31: i44–i52.
Moore, T. 1859. Report on the varieties of Achimenes. Proceedings of the Royal
Horticultural Society London 1: 454–464.
2018, Volume 105 Roberts and Roalson—Phylogenomics in Achimenes15
Moyle, L. C., M. S. Olson, and P. Tin. 2004. Patterns of reproductive isolation
in three angiosperm genera. Evolution 58: 1195–1208.
Payseur, B. A., and L. H. Rieseberg. 2016. A genomic perspective on hybridiza-
tion and speciation. Molecular Ecology 25: 2337–2360.
Perret, M., A. Chautems, R. Spichiger, T. G. Barraclough, and V. Savolainen.
2007. e geographical pattern of speciation and oral diversication in the
Neotropics: the tribe Sinningieae (Gesneriaceae) as a case study. Evolution
62: 1641–1660.
Petit, R. J., C. Bodénès, A. Ducousso, G. Roussel, and A. Kremer. 2004. Hybridization
as a mechanism of invasion in oaks. New Phytologist 161: 151–164.
Price, M. N., P. S. Dehal, and A. P. Arkin. 2010. FastTree 2—Approximately
maximum- likelihood trees for large alignments. PLoS ONE 5: e9490.
Ramírez Roa, M.A. 1987. Revision de Achimenes (Gesneriaceae). Ph.D. disser-
tation, Universidad Nacional Autonoma de Mexico, Mexico City, Mexico.
Ramsey, J., H. D. Bradshaw, and D. W. Schemske. 2003. Components of repro-
ductive isolation between the monkeyowers Mimulus lewisii and M. cardi-
nalis (Phrymaceae). Evolution 57: 1520–1534.
Roalson, E. H., J. K. Boggan, L. E. Skog, and E. A. Zimmer. 2005. Untangling
Gloxinieae (Gesneriaceae). I. Phylogenetic patterns and generic boundaries
inferred from nuclear, chloroplast, and morphological cladistic datasets.
Taxon 54: 389–410.
Roalson, E. H., and W. R. Roberts. 2016. Distinct processes drive diversication
in dierent clades of Gesneriaceae. Systematic Biology 65: 662–684.
Roalson, E. H., L. E. Skog, and E. A. Zimmer. 2003. Phylogenetic relation-
ships and the diversication of oral form in Achimenes (Gesneriaceae).
Systematic Botany 28: 593–608.
Roalson, E. H., L. E. Skog, and E. A. Zimmer. 2008. Untangling Gloxinieae
(Gesneriaceae). II. Reconstructing biogeographic patterns and estimat-
ing divergence times among New World continental and island lineages.
Systematic Botany 33: 159–175.
Roberts, W. R., and E. H. Roalson. 2017. Comparative transcriptome analyses
of ower development in four species of Achimenes (Gesneriaceae). BMC
Genomics 18: 240.
Roda, F., F. K. Mendes, M. W. Hahn, and R. Hopkins. 2017. Genomic evidence
for gene ow during reinforcement in Texas Phlox. Molecular Ecology 26:
Rosenzweig, B. K., J. B. Pease, N. J. Besansky, and M. W. Hahn. 2016. Powerful
methods for detecting introgressed regions from population genomic data.
Molecular Ecology 25: 2387–2397.
Savolainen, V., M.-C. Anstett, C. Lexer, I. Hutton, J. J. Clarkson, M. V. Norup,
M. P. Powell, et al. 2006. Sympatric speciation in palms on an oceanic island.
Nature 441: 210.
Schmidt-Lebuhn, A. N., B. Keller, J. M. De Vos, and E. Conti. 2012. Phylogenetic
analysis of Primula section Primula reveals rampant non- monophyly among
morphologically distinct species. Molecular Phylogenetics and Evolution 65:
Scopece, G., A. Muscacchio, A. Widmer, and S. Cozzolino. 2007. Patterns of
reproductive isolation in Mediterranean deceptive orchids. Evolution 61:
Servedio, M. R., and M. A. F. Noor. 2003. e role of reinforcement in speciation:
theory and data. Annual Review of Ecology, Evolution, and Systematics 34:
Smith, S. A., and C. W. Dunn. 2008. Phyutility: a phyloinformatics tool for trees,
alignments, and molecular data. Bioinformatics 24: 715–716.
Solís-Lemus, C., and C. Ané. 2016. Inferring phylogenetic networks with maxi-
mum pseudolikelihood under incomplete lineage sorting. PLoS Genetics 12:
Solís-Lemus, C., P. Bastide, and C. Ané. 2017. PhyloNetworks: a package for phy-
logenetic networks. Molecular Biology and Evolution 34: 3292–3298.
Solís-Lemus, C., M. Yang, and C. Ané. 2016. Inconsistency of species tree meth-
ods under gene ow. Systematic Biology 65: 843–851.
Soliva, M., and A. Widmer. 1999. Genetic and oral divergence among sym-
patric populations of Gymnadenia conopsea sl (Orchidaceae) with dierent
owering phenology. International Journal of Plant Sciences 160: 897–905.
Soliva, M., and A. Widmer. 2003. Gene ow across species boundaries in sym-
patric, sexually deceptive Ophrys (Orchidaceae) species. Evolution 57:
Stamatakis, A. 2014. RAxML version 8: a tool for phy logenetic analysis and post-
analysis of large phylogenies. Bioinformatics 30: 1312–1313.
Stebbins, G. L. 1950. Variation and evolution in plants. Columbia University
Press, New York.
Suyama, M., D. Torrents, and P. Bork. 2006. PAL2NAL: robust conversion of
protein sequence alignments into the corresponding codon alignments.
Nucleic Acids Research 34 (Suppl 2): W609–612.
Tin, P., M. S. Olson, and M. C. Moyle. 2001. Asymmetrical crossing barriers in
angiosperms. Proceedings of the Royal Society B 268: 861–867.
Vachaspati, P., and T. Warnow. 2015. ASTRID: accurate species trees from inter-
node distances. BMC Genomics 16: S3.
Vallejo-Marín, M., and S. J. Hiscock. 2016. Hybridization and hybrid speciation
under global change. New Phytologist 211: 1170–1187.
Vargas, O. M., E. M. Ortiz, and B. B. Simpson. 2017. Conicting phylogenomic
signals reveal a pattern of reticulate evolution in a recent high- Andean di-
versication (Asteraceae: Astereae: Diplostephium). New Phytologist 214:
Vereecken, N. J., S. Cozzolino, and F. P. Schiestl. 2010. Hybrid oral scent nov-
elty drives pollinator shis in sexually deceptive orchids. BMC Evolutionary
Biology 10: 103.
Ward, B. J., and C. Oosterhout. 2015. HYBRIDCHECK: soware for the rapid
detection, visualization and dating of recombinant regions in genome se-
quence data. Molecular Ecology Resources 16: 534–539.
Wessinger, C. A., and M. D. Rausher. 2014. Predictability and irreversibility of
genetic changes associated with ower color evolution in Penstemon barba-
tus. Evolution 68: 1058–1070.
Whitney, K. D., R. A. Randell, and L. H. Rieseberg. 2006. Adaptive introgression
of herbivore resistance traits in the weedy sunower Helianthus annuus. The
American Naturalist 167: 794–807.
Wiehler, H. 1983. A synopsis of the neotropical Gesneriaceae. Selbyana. 6:
Wiehler, H. 1992. New species of Gesneriaceae from the Neotropics. Phytologia
73: 220–241.
Wolf, P. G., D. R. Campbell, N. M. Waser, S. D. Sipes, T. R. Toler, and J. K.
Archibald. 2001. Tests of pre- and postpollination barriers to hybridization
between sympatric species of Ipomopsis (Polemoniaceae). American Journal
of Botany 88: 213–219.
Yang, Y., and S. A. Smith. 2014. Orthology inference in nonmodel organisms
using transcriptomes and low- coverage genomes: improving accuracy and
matrix occupancy for phylogenomics. Molecular Biology and Evolution 31:
APPENDIX 1. Taxa used for the analyses; Collector; Voucher number; WS = Washington State University.
Achimenes admirabilis Wiehler; W.R. Roberts; WR0569; WS. Achimenes antirrhina (DC.) C.V. Morton; W.R. Roberts; WR0570; WS. Achimenes candida Lindl.;
W.R. Roberts; WR0571; WS. Achimenes cettoana H.E. Moore; W.R. Roberts; WR0155; WS. Achimenes erecta (Lam.) H.P. Fuchs; W.R. Roberts; WR0156; WS.
Achimenes grandiora (Schltdl.) DC.; W.R. Rober ts; WR0572; WS. Achimenes longiora DC.; W.R. Roberts; WR0573; WS. Achimenes misera Lindl.; W.R. Roberts;
WR0157; WS. Achimenes patens Benth.; W.R. Roberts; WR0158; WS. Achimenes pedunculata Benth.; W.R. Roberts; WR0574; WS. Eucodonia verticillata (M.
Martens & Galeotti) Wiehler; W.R. Roberts; WR0575; WS. Gesneria cuneifolia (DC.) Fritsch; W.R. Roberts; WR0576; WS.
... We previously inferred a phylogeny for the 12 sampled species using 1,306 single-copy orthologs identified from the same transcriptome dataset used here (Roberts & Roalson, 2018). For comparative analyses of module-trait relationships, we randomly sampled 50 single-copy ortholog gene trees and rescaled branch lengths to be proportional to time (ultrametric) using the "chronos" function in the R package ape (Paradis, Claude & Strimmer, 2004). ...
... For comparative analyses of module-trait relationships, we randomly sampled 50 single-copy ortholog gene trees and rescaled branch lengths to be proportional to time (ultrametric) using the "chronos" function in the R package ape (Paradis, Claude & Strimmer, 2004). Bootstrap support was 100 for nearly every branch in the Roberts & Roalson (2018) phylogeny, therefore we chose to use randomly sampled ortholog trees to account for phylogenetic uncertainty. ...
... relationships and floral traits in Achimenes. Phylogeny of Achimenes adapted fromRoberts & Roalson (2018). Clade 1 and Clade 2 (sensuRoalson, Skog & Zimmer (2003)) are indicated on the phylogeny with a circle and box, respectively. ...
Full-text available
Background Genetic pathways involved with flower color and shape are thought to play an important role in the development of flowers associated with different pollination syndromes, such as those associated with bee, butterfly, or hummingbird pollination. Because pollination syndromes are complex traits that are orchestrated by multiple genes and pathways, the gene regulatory networks have not been explored. Gene co-expression networks provide a systems level approach to identify important contributors to floral diversification. Methods RNA-sequencing was used to assay gene expression across two stages of flower development (an early bud and an intermediate stage) in 10 species of Achimenes (Gesneriaceae). Two stage-specific co-expression networks were created from 9,503 orthologs and analyzed to identify module hubs and the network periphery. Module association with bee, butterfly, and hummingbird pollination syndromes was tested using phylogenetic mixed models. The relationship between network connectivity and evolutionary rates ( dN / dS ) was tested using linear models. Results Networks contained 65 and 62 modules that were largely preserved between developmental stages and contained few stage-specific modules. Over a third of the modules in both networks were associated with flower color, shape, and pollination syndrome. Within these modules, several hub nodes were identified that related to the production of anthocyanin and carotenoid pigments and the development of flower shape. Evolutionary rates were decreased in highly connected genes and elevated in peripheral genes. Discussion This study aids in the understanding of the genetic architecture and network properties underlying the development of floral form and provides valuable candidate modules and genes for future studies.
... Many other studies suggested insect-mediated pollination and generalist pollination, and bees are cited many times as pollination in the hybridizing species (Nettel et al., 2008;Scotti-Saintagne et al., 2013;Luebert et al., 2014;Mori, Zucchi & Souza, 2015;Baena-Díaz et al., 2018;Nevado et al., 2018;Tapia-Pastrana, 2020). For example, André et al. (2022) Roberts & Roalson (2018) studied species of Gesneriaceae and reported bees as pollinators in addition to hummingbirds and butterflies and suggested bees and butterflies as pollen vectors between species. Other studies also reported generalist pollinators, as in species of Begonia L. (Begoniaceae; Twyford, Kidner & Ennos, 2015); and species of Brahea Mart. ...
Coevolution between floral traits and specific pollination behaviour is a significant evolutionary force in angiosperm diversification. However, hybridization is also reported to occur between plants with specialist pollination syndromes. Understanding the role of pollinators in plant diversification is crucial, especially in megadiverse regions, such as the Neotropics. In this review, we examine plant hybridization studies in the Neotropics with the aim of providing a perspective on biotic and abiotic factors starting hybrid zone formation. The Pleistocene was the most widely cited time for the occurrence of hybridization facilitated by geographical range shifts, but time-calibrated analyses are needed to recover a more realistic scenario. Our synthesis of hybridization and pollination systems suggested that specialist and generalist pollinators were associated with the starting point of hybridization in the Neotropical flora. Bees and hummingbirds are most likely to be the primary vectors of interspecific gene flow, and even sporadic visits by bees or other generalist pollinators could allow the formation of a new hybrid zone. We highlight that seed and pollen dispersal vectors should be included in an integrative discussion on hybridization in the Neotropical flora. We also provide a preliminary map of hybrid zones in the Neotropics, including Brazilian vegetation cover and losses in the last 30 years, with the aim of encouraging research into human-driven anthropogenic changes and formation and/or shift of hybrid zones through time.
... Currently, there are several lines of evidence for speciation with gene flow in plants (e.g., Muniz et al., 2022;Papadopulos et al., 2011Papadopulos et al., , 2019Roberts & Roalson, 2018) as well as in animals (e.g., Camurugi et al., 2021;Kautt et al., 2016Kautt et al., , 2020Martin et al., 2013;Niemiller et al., 2008;Oliveira et al., 2015;Roux et al., 2016). One prediction of this model is that the level of divergence should be heterogeneous across the genome, given that alleles at some loci are likely to be shared between incipient species, while selection maintains divergence at other loci Nosil et al., 2009;Riesch et al., 2017;Seehausen et al., 2014;Turner et al., 2005;Wu, 2001). ...
Full-text available
The study of mechanisms that generate new species is considered fundamental for broad areas of ecology and evolution. Speciation is a continuous process in which reproductive isolation is established, and it is of fundamental importance to understand the origins of the adaptations that contribute to this process. Hybrid zones are considered natural laboratories for the study of speciation and represent ideal systems for such studies. Here, we investigated genomic differentiation between hybridizing Neotropical species Pitcairnia staminea (G. Lodd.) and P. albiflos (Herb.). Using thousands of SNPs genotyped through RAD‐seq, we estimate effective population sizes, interspecific gene flow, as well as time of divergence between these two sister species and identify candidate genomic regions for positive selection that may be related to reproductive isolation. We selected different scenarios of speciation and tested them by using approximate Bayesian computation (ABC); we found evidence of divergence with gradual reduction in gene flow between these species over time, compatible with the hypothesis of speciation with gene flow between these Pitcairnia species. The parameter estimates obtained through ABC suggested that the effective population size of P. albiflos was around three times larger than that of P. staminea. Our divergence date estimates showed that these two species diverged during the Pliocene (4.7 Mya; CI = 1.3–8.5 Mya), which has likely allowed this species to accumulate genome‐wide differences. We also detected a total of 17 of 4165 loci which showed signatures of selection with high genetic differentiation (FST > 0.85), 12 of these loci were annotated in de novo assembled transcriptomes of both species, and 4 candidate genes were identified to be putatively involved in reproductive isolation. These four candidate genes were previously associated with the function of pollen development, pollen tube germination and orientation, abiotic stress, and flower scent in plants, suggesting an interplay between pre‐ and postpollination barriers in the evolution of reproductive isolation between such species. Graphic representation of the simulated scenarios of divergence between Pitcairnia albiflos and P. staminea tested using approximate Bayesian computation (ABC) framework. By contrasting different scenarios of speciation using ABC, we found evidence of divergence with gradual reduction in gene flow between these species over time, compatible with the hypothesis of speciation with gene flow between these Neotropical plant species.
... Among herbaceous plants found in montane tropical forests, there are even more significant evolutionary outcomes from hybridization. Recent transcriptomic work by Roberts & Roalson (2018) uncovered hybridization between ten of 24 species of Achimenes Pers. (Gesneriaceae), coupled with introgression among species pairs that share pollinators and produce viable hybrid offspring, suggesting that novel hybrid lineages may persist in these environments. ...
Full-text available
Hybridisation can facilitate both evolutionary diversification and extinction and has had a critical role in plant evolution, with~ 25% of species known to hybridise in some temperate floras. However, in the species-rich neotropical flora, hybridisation’s role in the evolution of diversity remains unclear. Our review examines studies of hybridisation in seed plants from across the Neotropics, and explores its outcomes on neotropical plant evolution. We review studies on a per-biome basis and a spectrum of evolutionary outcomes from hybridisation are evident across neotropical biomes and taxa. These range from short-term impacts, such as the broadening of ecological amplitude in hybrid progeny with transgressive phenotypes and genetic swamping, through to long term impacts such as the generation of new lineages. Among these studies certain themes emerge, such as the pervasive hybridisation among species-rich plant radiations from the Andean Páramos, suggesting a role for hybridisation in rapid diversification events. Finally, we highlight that hybridisation is relatively understudied in the neotropical flora despite its remarkable species-richness. The advent of genomic techniques can facilitate the study of hybridisation and its effects in understudied biomes and plant groups. The increasing availability of genomic resources will eventually allow comparisons between tropical and temperate floras and therefore shed light on hybridisation’s evolutionary impacts across the latitudinal biodiversity gradient.
... The application of genomic data has become an effective means to validate hybridization events. Many natural hybridization events have been proposed and validated in numerous organisms (Roberts and Roalson, 2018;Cao et al., 2019;Glémin et al., 2019;Wang et al., 2019;Zhang et al., 2019;Yang et al., 2020). In this study, our nuclear and cp genome analyses confirmed hybrid origin of C. × pinnata between C. brevicaudata and C. heracleifolia/C. ...
Full-text available
Interspecific hybridization is common and has often been viewed as a driving force of plant diversity. However, it raises taxonomic problems and thus impacts biodiversity estimation and biological conservation. Although previous molecular phylogenetic studies suggested that interspecific hybridization may be rather common in Clematis, and artificial hybridization has been widely applied to produce new Clematis cultivars for nearly two centuries, the issue of natural hybridization of Clematis has never been addressed in detail. In this study, we tested the hybrid origin of a mesophytic and cold-adapted vine species, Clematis pinnata, which is a rare and taxonomically controversial taxon endemic to northern China. Using field investigations, flow cytometry (FCM), phylogenomic analysis, morphological statistics, and niche modeling, we tested hybrid origin and species status of C. pinnata. The FCM results showed that all the tested species were homoploid (2n = 16). Phylonet and HyDe analyses based on transcriptome data showed the hybrid origins of C. × pinnata from either C. brevicaudata × C. heracleifolia or C. brevicaudata × C. tubulosa. The plastome phylogeny depicted that C. × pinnata in different sampling sites originated by different hybridization events. Morphological analysis showed intermediacy of C. × pinnata between its putative parental species in many qualitative and quantitative characters. Niche modeling results suggested that C. × pinnata had not been adapted to a novel ecological niche independent of its putative parents. These findings demonstrated that plants of C. × pinnata did not formed a self-evolved clade and should not be treated as a species. The present study also suggests that interspecific hybridization is a common mechanism in Clematis to generate diversity and variation, and it may play an important role in the evolution and diversification of this genus. Our study implies that morphological diversity caused by natural hybridization may overstate the real species diversity in Clematis.
... For example, relationships among the major lineages of Gesneriaceae are still poorly resolved and there is still no firm consensus on the phylogenetic positions of taxa such as Peltanthera, Sanango, Titanotrichum and Calceolariaceae (Weber et al., 2013;APG IV, 2016). Recent phylogenomic approaches provide the opportunity to fill these gaps in the Gesneriaceae, but so far they have been applied in few groups to solve issues of incomplete lineage sorting and hybridization (in Achimenes: Roberts and Roalson, 2018; in Cyrtandra: Kleinkopf et al., 2019). In the present study, we developed a gene capture method for sequencing hundreds of nuclear genes simultaneously and evaluated the utility of this dataset for phylogenetic studies both at deep and shallow evolutionary levels within the Gesneriaceae. ...
Full-text available
Gesneriaceae (ca. 3400 species) is a pantropical plant family with a wide range of growth form and floral morphology that are associated with repeated adaptations to different environments and pollinators. Although Gesneriaceae systematics has been largely improved by the use of Sanger sequencing data, our understanding of the evolutionary history of the group is still far from complete due to the limited number of informative characters provided by this type of data. To overcome this limitation, we developed here a Gesneriaceae-specific gene capture kit targeting 830 single-copy loci (776,754 bp in total), including 279 genes from the Universal Angiosperms-353 kit. With an average of 557,600 reads and 87.8% gene recovery, our target capture was successful across the family Gesneriaceae and also in other families of Lamiales. From our bait set, we selected the most informative 418 loci to resolve phylogenetic relationships across the entire Gesneriaceae family using maximum likelihood and coalescent-based methods. Upon testing the phylogenetic performance of our baits on 78 taxa representing 20 out of 24 subtribes within the family, we showed that our data provided high support for the phylogenetic relationships among the major lineages, and were able to provide high resolution within more recent radiations. Overall, the molecular resources we developed here open new perspectives for the study of Gesneriaceae phylogeny at different taxonomical levels and the identification of the factors underlying the diversification of this plant group.
... The phylogenetic network analysis suggests at least three exchanges of genetic material have occurred between unrelated lineages, several of which occur along deeper (internal) edges of the phylogeny. Estimates of the amounts of the genome exchanged between lineages during these events are high (26, 33, and 47%), especially when considering animal systems (Solís-Lemus and Ané 2016;Blair et al. 2019;Morando et al. 2020;Pyron et al. 2020) but are not atypical for plant systems (Crowl et al. 2017;Morales-Briones et al. 2018;Roberts and Roalson 2018). Given high CFs for sister relationships not analyzed by SNaQ (Fig. 1), it is likely that the P. carneum-P. ...
Phylogenomic data from a rapidly increasing number of studies provide new evidence for resolving relationships in recently radiated clades, but they also pose new challenges for inferring evolutionary histories. Most existing methods for reconstructing phylogenetic hypotheses rely solely on algorithms that only consider incomplete lineage sorting as a cause of intra- or inter-genomic discordance. Here, we utilize a variety of methods, including those to infer phylogenetic networks, to account for both incomplete lineage sorting and introgression as a cause for nuclear and cytoplasmic-nuclear discordance using phylogenomic data from the recently radiated flowering plant genus Polemonium (Polemoniaceae), an ecologically diverse genus in Western North America with known and suspected gene flow between species. We find evidence for widespread discordance among nuclear loci that can be explained by both incomplete lineage sorting and reticulate evolution in the evolutionary history of Polemonium. Furthermore, the histories of organellar genomes show strong discordance with the inferred species tree from the nuclear genome. Discordance between the nuclear and plastid genome is not completely explained by incomplete lineage sorting, and only one case of discordance is explained by detected introgression events. Our results suggest that multiple processes have been involved in the evolutionary history of Polemonium and that the plastid genome does not accurately reflect species relationships. We discuss several potential causes for this cytoplasmic-nuclear discordance, which emerging evidence suggests is more widespread across the Tree of Life than previously thought.
... According to the phylogenetic analysis by Roberts and Roalson (2018), Achimenes is a genus of Mesoamerican origin that comprises approximately 26 species. Achimenes antirrhina, A. flava, and A. patens belong to a larger clade comprising 10 species, and although these species are not supported as sister to one another, they last shared a common ancestor approximately 4 million years ago (Ma) (Roalson and Roberts, 2016); therefore, they are species of recent divergence. ...
Full-text available
Premise: Closely related species occurring in sympatry may experience the negative consequences of interspecific pollen transfer if reproductive isolation (RI) barriers are not in place. We evaluated the importance of pre- and post-pollination RI barriers in three sympatric species of Achimenes (Gesneriaceae), including ecogeographic, phenological, floral isolation, self-pollination, and hybrid viability (fruit and seed set). Methods: We recorded geographic distribution throughout species ranges and assessed flowering phenology and pollinator visitation at one site in central Mexico. In the greenhouse, we measured floral traits involved in RI and quantified fruit and seed set for from self, intraspecific, and interspecific crosses. Results: Ecogeographic barriers were important in RI, but under sympatry, phenological and floral barriers contributed more to total RI. Phenological RI varied between species and years, while floral RI was 100% effective at preventing interspecific visitation. Species showed differences in floral morphology, color, and scents associated with specialized pollination systems (A. antirrhina-hummingbirds, A. flava-bees, A. patens-butterflies); heterospecific visitation events were restricted to rare secondary pollinators. Hybrid crosses consistently yielded progeny in lower numbers than intraspecific crosses. Conclusions: This study indicated that neither autogamy nor early post-pollination barriers prevent interspecific pollen flow between Achimenes species. However, floral isolation, acting through a combination of attraction and reward traits, consistently ensures specificity of the pollination system. These results suggest that selection on floral traits to reduce the costs of hybrid progeny production may have played a role in evolution or maintenance of specialized pollination systems in Achimenes.
... A multitude of different HTS techniques have been used to resolve difficult groups and understand how their evolutionary histories have been affected by hybridization and/or ILS. Restriction-site associated DNA sequencing (RAD-seq; Hipp et al., 2014;Vargas et al., 2017), transcriptomics (Yang & Smith, 2014;Roberts & Roalson, 2018), genome skimming (Straub et al., 2012;Bock et al., 2014;Weitemier et al., 2014), and targeted enrichment Stull et al., 2013;Weitemier et al., 2014;Folk et al., 2015Folk et al., , 2017, among other methods, have been used in phylogenomic studies. Targeted enrichment is among the leading strategies for obtaining large amounts of data for relatively low cost, which has yielded highly informative data for resolving difficult groups, particularly those under putative ILS and/or hybridization complexity (Asclepias, Weitemier et al., 2014;Heuchera, Folk et al., 2017). ...
Full-text available
Cyrtandra (Gesneriaceae) is a genus of flowering plants with over 800 species distributed throughout Southeast Asia and the Pacific Islands. On the Hawaiian Islands, 60 named species and over 89 putative hybrids exist, most of which are identified on the basis of morphology. Despite many previous studies on the Hawaiian Cyrtandra lineage, questions regarding the reconciliation of morphology and genetics remain, many of which can be attributed to the relatively young age and evidence of hybridization between species. We utilized targeted enrichment, high‐throughput sequencing, and modern phylogenomics tools to test 31 Hawaiian Cyrtandra samples (22 species, two putative hybrids, four species with two samples each, one species with four samples) and two outgroups for species relationships and hybridization in the presence of incomplete lineage sorting (ILS). Both concatenated and species‐tree methods were used to reconstruct species relationships, and network analyses were conducted to test for hybridization. We expected to see high levels of ILS and putative hybrids intermediate to their parent species. Phylogenies reconstructed from the concatenated and species‐tree methods were highly incongruent, most likely due to high levels of incomplete lineage sorting. Network analyses inferred gene flow within this lineage, but not always between taxa that we expected. Multiple hybridizations were inferred, but many were on deeper branches of the island lineages suggesting a long history of hybridization. We demonstrated the utility of high‐throughput sequencing and a phylogenomic approach using 569 loci to understanding species relationships and gene flow in the presence of ILS. This article is protected by copyright. All rights reserved.
Full-text available
The phylogeny of eutherian mammals contains some of the most recalcitrant nodes in the tetrapod tree of life. We combined comprehensive taxon and character sampling to explore three of the most debated interordinal relationships among placental mammals. We performed in silico extraction of ultraconserved element loci from 72 published genomes and invitro enrichment and sequencing of ultraconserved elements from 28 additional mammals, resulting in alignments of 3,787 loci. We analyzed these data using concatenated and multispecies coalescent phylogenetic approaches, topological tests, and exploration of support among individual loci to identify the root of Eutheria and the sister groups of tree shrews (Scandentia) and horses (Perissodactyla). Individual loci provided weak, but often consistent support for topological hypotheses. Although many gene trees lacked accepted species-tree relationships, summary coalescent topologies were largely consistent with inferences from concatenation. At the root of Eutheria, we identified consistent support for a sister relationship between Xenarthra and Afrotheria (i.e., Atlantogenata). At the other nodes of interest, support was less consistent. We suggest Scandentia is the sister of Primatomorpha (Euarchonta), but we failed to reject a sister relationship between Scandentia and Glires. Similarly, we suggest Perissodactyla is sister to Cetartiodactyla (Euungulata), but a sister relationship between Perissodactyla and Chiroptera remains plausible.
Full-text available
Background Flowers have an amazingly diverse display of colors and shapes, and these characteristics often vary significantly among closely related species. The evolution of diverse floral form can be thought of as an adaptive response to pollination and reproduction, but it can also be seen through the lens of morphological and developmental constraints. To explore these interactions, we use RNA-seq across species and development to investigate gene expression and sequence evolution as they relate to the evolution of the diverse flowers in a group of Neotropical plants native to Mexico—magic flowers (Achimenes, Gesneriaceae). ResultsThe assembled transcriptomes contain between 29,000 and 42,000 genes expressed during development. We combine sequence orthology and coexpression clustering with analyses of protein evolution to identify candidate genes for roles in floral form evolution. Over 25% of transcripts captured were distinctive to Achimenes and overrepresented by genes involved in transcription factor activity. Using a model-based clustering approach we find dynamic, temporal patterns of gene expression among species. Selection tests provide evidence of positive selection in several genes with roles in pigment production, flowering time, and morphology. Combining these approaches to explore genes related to flower color and flower shape, we find distinct patterns that correspond to transitions of floral form among Achimenes species. Conclusions The floral transcriptomes developed from four species of Achimenes provide insight into the mechanisms involved in the evolution of diverse floral form among closely related species with different pollinators. We identified several candidate genes that will serve as an important and useful resource for future research. High conservation of sequence structure, patterns of gene coexpression, and detection of positive selection acting on few genes suggests that large phenotypic differences in floral form may be caused by genetic differences in a small set of genes. Our characterized floral transcriptomes provided here should facilitate further analyses into the genomics of flower development and the mechanisms underlying the evolution of diverse flowers in Achimenes and other Neotropical Gesneriaceae.
Full-text available
Gene flow can impede the evolution of reproductive isolating barriers between species. Reinforcement is the process by which pre-zygotic reproductive isolation evolves in sympatry due to selection to decrease costly hybridization. It is known that reinforcement can be prevented by too much gene flow, but we still do not know how often have pre-zygotic barriers evolved in the presence of gene flow or how much gene flow can occur during reinforcement. Flower color divergence in the native Texas wildflower, Phlox drummondii, is one of the best-studied cases of reinforcement. Here we use genomic analyses to infer gene flow between P. drummondii and a closely related sympatric species, P. cuspidata. We de novo assemble transcriptomes of four Phlox species to determine the phylogenetic relationships between these species, and find extensive discordance among gene tree topologies across genes. We find evidence of introgression between sympatric P. drummondii and P. cuspidata using the D-statistic, and use phylogenetic analyses to infer the predominant direction of introgression. We investigate geographic variation in gene flow by comparing the relative divergence of genes displaying discordant gene trees between an allopatric and sympatric sample. These analyses support the hypothesis that sympatric P. drummondii has experienced gene flow with P. cuspidata. We find that gene flow between these species is asymmetrical, which could explain why reinforcement caused divergence in only one of the sympatric species. Given the previous research in this system, we suggest strong selection can explain how reinforcement successfully evolved in this system despite gene flow in sympatry. This article is protected by copyright. All rights reserved.
Full-text available
The role of hybridization in adaptive evolution is contentious. While many cases of adaptive trait introgression have been proposed, the relevant traits have rarely been identified, resulting in a lack of clear examples of this process. Here, we examine a purported case of adaptive introgression in which the annual sunflower Helianthus annuus annuus has captured alleles from a congener (Helianthus debilis) to form a stabilized hybrid, Helianthus annuus texanus. We tested the hypotheses that herbivore resistance traits have introgressed from H. debilis to H. annuus and have increased adaptation in the latter. In two common gardens, fitness (estimated by seed production) was on average 55% higher in H. a. texanus than in H. a. annuus. For H. a. texanus, three damage traits (of seven tested) differed significantly from the H. a. annuus parent in one or both sites and were shifted in the direction of the more resistant H. debilis. Natural selection favored H. a. \documentclass{aastex} \usepackage{amsbsy} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{bm} \usepackage{mathrsfs} \usepackage{pifont} \usepackage{stmaryrd} \usepackage{textcomp} \usepackage{portland,xspace} \usepackage{amsmath,amsxtra} \usepackage[OT2,OT1]{fontenc} \newcommand\cyr{ \renewcommand\rmdefault{wncyr} \renewcommand\sfdefault{wncyss} \renewcommand\encodingdefault{OT2} \normalfont \selectfont} \DeclareTextFontCommand{\textcyr}{\cyr} \pagestyle{empty} \DeclareMathSizes{10}{9}{7}{6} \begin{document} \landscape $annuus\times H.$ \end{document} debilis BC1 hybrids (synthesized to mimic the ancestors of H. a. texanus) with H. debilis–like resistance to seed midges Neolasioptera helianthis and to receptacle/seed feeding Lepidoptera at one or both sites. Assuming similar herbivore pressures in the past, these results suggest that introgression of biotic resistance traits was important in the adaptation of H. annuus to central and southern Texas.
PhyloNetworks is a Julia package for the inference, manipulation, visualization and use of phylogenetic networks in an interactive environment. Inference of phylogenetic networks is done with maximum pseudolikelihood from gene trees or multi-locus sequences (SNaQ), with possible bootstrap analysis. PhyloNetworks is the first software providing tools to summarize a set of networks (from a bootstrap or posterior sample) with measures of tree edge support, hybrid edge support, and hybrid node support. Networks can be used for phylogenetic comparative analysis of continuous traits, to estimate ancestral states or do a phylogenetic regression. The software is available in open source and with documentation at
Pollinators that forage indiscriminately can transfer pollen from one species to another, reducing the amount that reaches conspecific flowers. I present evidence that the presence of another plant species visited by the same pollinators can also reduce pollen dispersal distances and outcrossing. This has the potential to influence gene flow and reproductive success. Pollen carryover and movement patterns were measured for the shared insect pollinators of Stellaria pubera and Claytonia virginica in North Carolina. Bee flies deposited similar amounts of Stellaria pollen on a series of pistillate Claytonia flowers as on a series of pistillate Stellaria flowers. In arrays of potted plants, flies and solitary bees visited most flowers on a plant before leaving and then flew to a nearby plant chosen independently of species; 95% of moves were to one of 12 nearest neighbors. These measures of pollen carryover and movement patterns were used in a set of computer simulations to predict pollen dispersal distances. The simulations suggested that C. virginica substantially reduces outcrossing and pollen flow in S. pubera. These predictions were tested by tracking dye movement from anthers in populations of potted plants. Addition of C. virginica reduced the mean squared distance moved by dye to receptive S. pubera flowers by 23% and reduced the amount of dye moved by 51%. The estimated pollen component of gene flow was also much lower in a natural population of 5. pubera mixed with C. virginica than in the synthetic single-species populations.
High-throughput sequencing is helping biologists to overcome the difficulties of inferring the phylogenies of recently diverged taxa. The present study analyzes the phylogenetic signal of genomic regions with different inheritance patterns using genome skimming and ddRAD-seq in a species-rich Andean genus (Diplostephium) and its allies. We analyzed the complete nuclear ribosomal cistron, the complete chloroplast genome, a partial mitochondrial genome, and a nuclear-ddRAD matrix separately with phylogenetic methods. We applied several approaches to understand the causes of incongruence among datasets, including simulations and the detection of introgression using the D-statistic (ABBA-BABA test). We found significant incongruence among the nuclear, chloroplast, and mitochondrial phylogenies. The strong signal of hybridization found by simulations and the D-statistic among genera and inside the main clades of Diplostephium indicate reticulate evolution as a main cause of phylogenetic incongruence. Our results add evidence for a major role of reticulate evolution in events of rapid diversification. Hybridization and introgression confound chloroplast and mitochondrial phylogenies in relation to the species tree as a result of the uniparental inheritance of these genomic regions. Practical implications regarding the prevalence of hybridization are discussed in relation to the phylogenetic method.
Hybridization has played an important role in the evolution of many lineages. With the growing availability of genomic tools and advancements in genomic analyses, it is becoming increasingly clear that gene flow between divergent taxa can generate new phenotypic diversity, allow for adaptation to novel environments, and contribute to speciation. Hybridization can have immediate phenotypic consequences through the expression of hybrid vigor. On longer evolutionary time scales, hybridization can lead to local adaption through the introgression of novel alleles and transgressive segregation and, in some cases, result in the formation of new hybrid species. Studying both the abundance and the evolutionary consequences of hybridization has deep historical roots in plant biology. Many of the hypotheses concerning how and why hybridization contributes to biological diversity currently being investigated were first proposed tens and even hundreds of years ago. In this Update, we discuss how new advancements in genomic and genetic tools are revolutionizing our ability to document the occurrence of and investigate the outcomes of hybridization in plants. © 2017 American Society of Plant Biologists. All rights reserved.